Skip to main content

Full text of "Punched Cards Their Applications To Science And Industry (second edition)"

See other formats









ofi£Mxsm 





PUNCHED CARDS 

THEIR APPLICATIONS TO SCIENCE AND INDUSTRY 

SECOND EDITION 


Edited By 

ROBERT S. CASEY 

W. A. Sheaffer Pen Co. 

Fort Madison, Iowa 

JAMES W. PERRY 

Center for Documentation and Communication Research 
Western Reserve University 
Cleveland, Ohio 

MADELINE M. BERRY 

National Science Foundation 
Washington, D. C. 

And 

ALLEN KENT 

Center for Documentation and Communication Research 
Western Reserve University 
Cleveland,'^Ohio 


REINHOLD PUBLISHING CORPORATION 

NEW YORK 

CHAPMAN & HALL, LTD., LONDON 



Copyright 1958 by 

REINHOLD PUBLISHING CORPORATION 


All rights reserved 


Library of Congress Catalog Card Number: 58-12824 


REINHOLD PUBLISHING CORPORATION 
Publishers of Chemical Engineering Catalog, Chemical 
Materials Catalog, “Automatic Control/ 1 “Materials in 
Design Engineering,” “Progressive Architecture”; Ad¬ 
vertising Management of the American Chemical Society 


PRINTED IN THE U.8.A. 



X 

r / 

r > 




CHEMISTRY 


PREFACE TO THE SECOND EDITION UBaARI 


The editors are pleased with the reception given the first edition of this 
book. Our readers were apparently more tolerant of its shortcomings than 
were the editors. 

For this edition the purposes remain the same as stated in the preface to 
the first edition, namely: to help the individual worker, to record present 
knowledge and experience, and to discuss some general principles as a 
stimulus to further development. 

Some idea of the rapidity with which the field has grown may be gained 
from the fact that the bibliography of uses contains 400 entries, compared 
with 276 entries in the first edition. This great increase is reflected in the 
extension of the Practical Applications Section (Part II) from 186 pages 
in the first edition to 295 pages in the present book. Here the reader will 
find a broad survey of such important and unique uses as the Peek-a-Boo 
System; the Uniterm System; mechanized coding and searching techniques 
applied to the metallurgical literature; the Zato-coding System; and a most 
interesting discussion of the use of punched cards in linguistic analysis as 
applied to ancient texts such as the Dead Sea Scrolls. 

The general plan of presentation for Parts I to V is the same as in the 
first edition. The following chapters are unchanged: 4. Preparing Reports, 
Papers and Books; 20. Correlation of Research Data; 21. Mathematical 
Analysis of Coding; 26. Searching the Literature; 17. Plant Breeding and 
Genetics has been lengthened slightly. The other chapters were rewritten 
or replaced. 

The chapter on Computation was eliminated. The potentialities of 
punched-card machines for scientific computation are so vast that they 
cannot be covered, or even adequately outlined within the limitations of 
a chapter. The bibliography, Chapter 30, contains references to applica¬ 
tions of punched-card computations. 

Substantial advances in the science and art of punched-card applications 
have been recorded since the first edition. Subject matter analysis has been 
receiving the attention of numerous workers, and much progress has been 
reported. This is perhaps the most important topic in the punched-card 
field and like most other topics in that field, is still in need of further de¬ 
velopment. In certain limited areas advancements in the punched-card art 
are becoming more definitive; for example: Qualitative Analysis by Spectral 
Methods; Metallurgical Literature; Classification, Searching and Mechani¬ 
zation in the U. S. Patent Office; and Library Applications. 

Chapters 7 and 15 are representative case histories of installations of 
commercial systems which have had successful applications. 

iii 


697 



IV 


PREFACE TO THE SECOND EDITION 


The applications of punched cards are too extensive to report all out¬ 
standing work. It is possible in a single book to pick only cases which are 
representative of different fields and different methods. To illustrate the 
diversity of applications, Chapter 14 was included, which gives miniature 
case histories in a variety of fields. 

For beginners, unfamiliar with punched cards, the following chapters 
are recommended as an introductory survey: Chapter 1; in Chapter 2, p. 
12, pp. 18-23 (Elementary Subject Matter Analysis and Coding), and pp. 
27-29 (Supplementary Notes on Hand-Sorted Cards); Chapter 4 illustrates 
the use of a simple bibliographic punched-card file. Skim through Chapter 
14, Review of Applications. 

The editors are grateful for all the help received in preparing this second 
edition. They are particularly grateful to the authors of the chapters, and 
to friends who called their attention to punched-card references. R. S. C. 
thanks his employer, the W. A. Sheaffer Pen Company, for the use of the 
company’s facilities, and his secretary, Mrs. Dorothy Billman, for typing 
and other help. 


Robert S. Casey 
James W. Perry 
Madeline M. Berry 
Allen Kent 


October 1,1958 



PREFACE TO THE FIRST EDITION 


In the present phase of our scientific age, a situation has developed in 
which research “publication has been extended far beyond our present 
ability to make real use of the record. The summation of human experience 
is being expanded at a prodigious rate and the means we use for threading 
through the consequent maze to the momentarily important item is the 
same as was used in the days of square-rigged ships.” 1 The very bulk of 
the rapidly expanding mass of scientific and technical information threatens 
to impair the usefulness of scientific investigation. 

The tendency of accumulations of scientific and technical information 
to become unwieldy is evident even in files of very modest size; even with 
small files of information, dissatisfaction has developed with the results 
obtained from conventional tools, such as ordinary file cards, classified re¬ 
port files, etc. It has been discovered that considerable improvement in 
speed and ease of locating information in files of modest size can be achieved 
by using punched cards of simple type, viz., edge-punched cards sold in the 
United States under the trademarks “Keysort,” “E-Z Sort” and “Rocket” 
and in England as “Paramount” or “Cope-Chat” cards. 

This book is directed principally to the needs of the individual scientist, 
engineer, or other technologist, whether in the laboratory, field, industrial 
plant, library, school or executive office. Our primary purpose is to furnish 
sufficient information to permit the application of punched-card tech¬ 
niques to individual problems. However, the present state of knowledge of 
this subject does not allow full definitive treatment. There are many 
scattered bits of information about punched-card applications, each indi¬ 
cating the value of mechanical aids to the solution of intellectual problems. 
But, this knowledge needs to be extended and correlated. Many of the 
procedures described are preliminary, tentative and experimental. 

Therefore, another purpose of this book is to record present knowledge 
and experience so that better use of the presently available punched-card 
devices, and design of devices better suited to practical needs, will be 
stimulated. In addition, some general principles are discussed which may 
also apply to types of mechanical devices not yet invented. 

The hand-sorted edge-punched cards are discussed in greater detail than 
the machine-sorted cards. In fact, one object has been to make the book 
serve as an operating instruction manual for the edge-punched cards. It is 
not possible to do the same for machine-sorted cards within the scope of 
this book. The editors feel that familiarity with the easily learned sorting 

1 Bush, Vannevar, Atlantic Monthly, 176 , 101-8 (July, 1945). 



VI 


PREFACE TO THE FIRST EDITION 


procedures and techniques of using hand-sorted cards will facilitate under¬ 
standing of the more elaborate sorting procedures possible with machine- 
type cards. The machine-sorted cards and the machines for manipulating 
them are described and illustrated. The emphasis is on their use. Various 
applications and case histories are reported. The full possibilities of their 
applications are so vast that it is not possible to set down in one place all 
operating instructions. Detailed directions for each application must be 
obtained through consultation with technicians from the manufacturers. 

The editors do not imply preference for one type of card over another: 
on the contrary, they hope that this book will help to evaluate the short¬ 
comings and the advantages of different types of punched cards for different 
purposes. 

Part I is introductory and elementary. It contains sufficient information 
to permit an individual to set up and use a simple punched-card file. 

Part II consists of case histories of punched-card applications carefully 
selected to show what has already been accomplished. The most effective 
utilization of punched-card methods is learned more readily from practice 
and experience than from rule and precept. For this reason, Part II forms 
the heart of this book. 

Part III is more general and theoretical in nature. Fundamental problems 
involved in applying punched-card techniques to intellectual activities, 
and vice versa, are discussed. Some of the chapters of Part III consider 
the general problem of organizing information quite apart from punched- 
card techniques. It is hoped that these chapters may contribute generally 
to the advance of the art of information analysis and also stimulate further 
investigation of the possibilities inherent in punched-card techniques. 

Part IV is a study, speculative in nature, as to the role that punched 
cards and related devices may eventually play in relationship to other 
methods for coping with information problems. 

It is not possible at present to define the realm of usefulness of punched- 
card techniques. It is, of course, obvious that punched-card techniques will 
supplement rather than supplant existing methods in handling information. 
No revolutionary schemes on a large scale are advocated at present. 

Part V is a bibliography on uses of punched cards in connection with 
scientific information. Papers on the subject are so widely scattered that a 
r6sum6 of this sort appeared advisable. The editors will appreciate having 
their attention directed to any pertinent papers that may have been 
overlooked. 

As the chapters have been written by different authors, a certain amount 
of overlapping and repetition is inevitable. The editors feel that this is not 
wholly undesirable in view of the present undeveloped state of the art. 
Furthermore, the editors have made no effort to reconcile differences in 



PREFACE TO THE FIRST EDITION vii 

opinion among the various authors. It is not always possible to be certain 
what is irreconcilable difference in opinions and what is merely difference 
in viewpoint. The editors feel that this treatment is to the advantage of 
the discriminating reader. 

The editors solicit suggestions for improvement of future editions. 

Editing this book has been a pleasant task, thanks, first of all, to the 
generous cooperation of the authors of the various individual chapters. 
Thanks are also due the American Chemical Society, whose Board of Direc¬ 
tors, through its Committee on Punched Cards, has supported and encour¬ 
aged the study of punched-card techniques for chemical information 
problems. We are grateful to the punched-card companies whose products 
are mentioned throughout the book, for the use of illustrations and for 
much helpful technical information. Miss Madeline M. Berry contributed 
skillful assistance in checking the manuscript and reading proof. We also 
wish to thank Miss Alice M. Perry, who typed most of the manuscript. 

Completion of our editorial task within a reasonable time would scarcely 
have been possible without financial support accorded one of us (J. W. P.) 
by the Carnegie Foundation through the Center for Scientific Aids to 
Learning at M. I. T. The other editor (R. S. C.) thanks his employer, the 
W. A. Sheaffer Pen Co., for use of the company’s facilities, and his secretary, 
Mrs. Dorothy Billman, for much typing and other assistance. 

James W. Perry 
Cambridge, Mass. 

Robert S. Casey 
Fort Madison, Iowa 


June, 1951 




CONTENTS 


Chapter Page 

Preface to Second Edition . iii 

Preface to First Edition . v 

PART I: FUNDAMENTAL MACHINE CONSIDERATIONS 

1. Introduction, Allen Kent , James W. Perry , and Robert S . Casey . 3 

2. Elementary Manipulations of Hand-Sorted Punched Cards, Robert S . 

Casey and James W. Perry . 12 

3. Commercially Available Equipment and Supplies, Thomas H. Rees t Jr.. 30 

PART II: PRACTICAL APPLICATIONS OF PUNCHED CARDS 
AND RELATED DEVICES 

4. The Use of Punched-Card Techniques in Preparing Reports, Papers, 

and Books, Charles A. Burkhard . 93 

5. An International Classification and Punched-Card Filing System for 

Metallurgical Literature, Marjorie R . Hyslop and Alvina Wassenberg. 100 

6. The Peek-a-Boo System—Optical Coincidence Subject Cards in Infor¬ 

mation Searching, W. A. Wildhack and Joshua Stern . 125 

7. A Uniterm System for Reports, Charles E. Zerwekh , Jr . 152 

8. An Improved Anesthesia Record Card, Max S. Sadove and Myron J. Levin. 161 

9. Punched Cards as Aids to Qualitative Chemical Analysis by Spectral 

Methods, L. E. Kuentzel . 175 

10. An Application of Random Codes for Literature Searching, Claire K. 

Schultz . 232 

11. Searching Metallurgical Literature, Allen Kent and James W. Perry.. 248 

12. Classification, Searching and Mechanization in the U. S. Patent 

Office, B. E. Lanham and J. Leibowitz . 261 

13. Application of Punched Cards to Library Routines, Madeline M. Berry. 279 

14. Review of Applications, Barbara L. Haksteen . 303 

15. A Case History of a Zatocoding Information Retrieval System, Claude 

W. Brenner and Calvin N . Mooers . 340 

16. The Use of Punched Cards in Linguistic Analysis, Rev . Roberto Busa. .. 357 

17. An Abstracting and Information Service for Plant Breeding and 

Genetics, R. H. Richens . 374 


IX 
















X 


CONTENTS 


PART III: FUNDAMENTAL CONSIDERATIONS 
IN CODING AND SYSTEMS DESIGN 

18. Subject Matter Analysis and Coding—Some Fundamental Considera¬ 

tions, James W. Perry . 391 

19. Holes, Punches, Notches, Slots, and Logic, C. D. Gull . 422 

20. Correlation of Research Data and Establishment of Cause and Effect 

Relationships, D . E . H. Frear . 432 

21. Mathematical Analysis of Coding Systems, Carl S. Wise . 438 

22. Comprehensive Coding Schemes for Chemical Compounds, D. E. H. 

Frear . 465 

23. Superimposed Coding with the Aid of Randomizing Squares for Use In 

Mechanical Information Searching Systems, H . P. Luhn . 492 

24. Indexing and Index Searching, E. J. Crane and Charles L. Bernier . 510 

25. Making Classification Systems for Punched-Card Coding, Norman T . 

Ball . 528 

26. Searching the Literature, Byron A. Soule . 542 

27. Transcription Problems in Preparing and Using Punched-Card Files, 

C. D. Gull . 555 

PART IV: FUTURE POSSIBILITIES 

28. Evaluation of Mechanized Documentation at the Gmelin Institute, 

E. H. Eric Pietsch . 571 

29. Data Processing Machines for Filing and Retrieving Technical In¬ 

formation, Ascher Opler . 619 

PART V: ANNOTATED BIBLIOGRAPHY ON 
USES OF PUNCHED CARDS 

30. Annotated Bibliography on Uses ok Punched Cards, Robert S. Casey. .. 638 

Author Index. 673 

Subject Index. 679 

















Part I 

FUNDAMENTAL MACHINE 
CONSIDERATIONS 




Chapter 1 

INTRODUCTION 


Allen Kent and James W. Perry 
Western Reserve University, Cleveland, Ohio 
AND 

Robert S. Casey 

W. A. Sheaffer Pen Co., Fort Madison, Iowa 

Methods and systems for expediting the recall and correlation of re¬ 
corded information by applying various mechanical and electronic devices 
have made rapid strides during the past decade. The “recorded informa¬ 
tion” may be entries in notebooks, records on pieces of paper, correspond¬ 
ence files, collections of reprints, notes on file cards, accounts, financial 
transactions, measurements, calculations, pictures, diagrams, drawings, 
descriptions of people and things—almost anything the human mind can 
conceive. “Recorded information” on one hand represents all the books 
and journals ever printed; on the other hand, it is the growing accumula¬ 
tion of data in your own laboratory or office. It is toward solution of the 
latter problem that this book is directed. 

Punched cards are being applied to a steadily widening range of subject 
matter. One result has been to stimulate interest in similar applications of 
various electronic devices, especially computers, and another has been to 
initiate the development of specially designed searching and selecting ma¬ 
chines. At present, however, a variety of punched cards is the most widely 
used type of mechanical aid for facilitating the retrieval and correlation of 
recorded information. 

The two general types of punched cards, for hand sorting and machine 
sorting, have been in the process of development for almost two centuries. 
The control card for looms, invented in 1780 by Joseph Jacquard, laid the 
foundation for the future development of information storage tools. The 
loom control card stored the information necessary to reproduce patterns 
consistently during the weaving of fabrics. 

Another pioneering development of unusual importance in the develop¬ 
ment of present-day information control devices was the “analytical en¬ 
gine” invented by Charles Babbage about 1840. This device used pre¬ 
punched cards to facilitate statistical control. 

The first punched cards and equipment for manipulating them which 
resembled modern counterparts appeared about 1880, when Dr. Herman 


3 



4 


PUNCHED CARDS 


Hollerith introduced a pantograph punch and an electric accounting tabu¬ 
lator with sorting box for use in connection with the United States Census. 
The commercial organization formed by Hollerith, The Tabulating Ma¬ 
chine Company, which merged in 1911 into the Computing-Tabulating- 
Recording Company, was the forerunner of International Business Ma¬ 
chines Corporation and paved the way for other companies entering this 
field (see chapter 3). 

Hand-sorted punched cards as an aid to “preventing the accidental mis¬ 
placement of a card in the files” made an elementary appearance in 1904 
with a system which required the notching of the bottom edges of cards 
according to the file section in which the cards were to be placed. Rods in 
a suitable holder at the bottom of the card tray forced incorrectly filed 
cards (notches in wrong place) to pop up when placed in any but the cor¬ 
rect section of the file. 1 Another invention in 1907, based upon a variation 
of this principle, had lifting-bars at the bottom of the card tray which could 
be inserted in appropriate positions in order to select the desired cards 
which were notched at various positions along the bottom edge of the card.* 
These developments appear to have been the forerunners of the hand- 
sorted punched cards commercially available today. 

The What, Why and How of using punched cards are not as obvious as 
they are in the case of wrenches, hammers, screw drivers and other simple 
tools. Some idea of the What and Why of punched cards is given in this 
chapter. 

As mentioned earlier, the two general types of punched cards in common 
use are hand-sorted and machine-sorted. The hand-sorted type has one, 
two or more rows of holes along one or more edges of the card. Meanings 
are assigned to individual holes or to combinations of holes. The holes on 
a given card, appropriate to the entries to be punched on that card, are 
clipped open to the edge of the card, forming notches as shown in Figure 1-1. 
The sorting needle or “tumbler” resembles a single-tine ice pick with a 
blunt point or a knitting needle with a handle. When it is inserted in a 
given hole in a group of cards and lifted, the cards on which that hole has 
been notched drop from the pack (Figure 1-2). 

Two different types of machine-sorted cards are illustrated in Figure 1-3. 
In the IBM card, the twelve punching positions in a vertical column con¬ 
stitute a coding unit. Digits are indicated by punching one of the positions 
number 0-9, while for each of the letters two holes are punched in the same 
column. In Remington-Rand cards, the twelve holes in each column are 
divided into two sets of six holes each. Each set of holes is a coding unit. 
In the punching system used by Remington-Rand, meaning is attached to 

1 U. S. Patent No. 759,483, to W. K. Sparrow (May 10, 1904). 

* U. S. Patent No. 873,305, to E. Eckart (Dec. 10, 1907). 



INTRODUCTION 


5 



Figure 1-1. Punching a notch in the edge of a hand-sorted card. 



Figure 1-2. Hand sorting a file of edge-punched cards. 


the punching of a single hole in a set and to the punching of combinations 
of two and three holes. For details see Figure 1-3. 

The cards are mechanically punched and in actual use are fed auto¬ 
matically through machines in which the punched holes cause electrical or 
mechanical contacts to be made, thus actuating mechanisms which per¬ 
form the desired operations. If desired, all cards coded in a given manner 


6 


PUNCHED CARDS 



0123HSC.7 89 


ABCBCFfiHI JKLMNOPQRSTUVWIIVZ 

••• ••• •••• 


ii'iiVt*ii V **•»*•» "■* "•* ’** "** '■* WUlf "uutfiiWWu 

““““““00“““““““““““00u“0000“““000““0“0*“O““““ 
““““““““00““““““09“0““9“999“09““9“9“9““““0hh“ 
“ 00 ’• iii» hhh “ “009999 “00 “0 ““ * 



Figure 1-3. Machine-sorted punched cards. Symbols, coded in vertical columns, 
are printed at top of card. 


can be selected from a file. Also, the file can be sorted into numerical or 
alphabetical order. The numerals or letters coded on the card can be printed 
on the card or on another form. Calculations can be made, coded cards 
reproduced, and numerous other operations carried out, all of which are 
controlled by the patterns of holes on the cards and by the use of the ap¬ 
propriate machines. Other systems, proposed or in various stages of de¬ 
velopment, use jets of air, light rays, or fluorescent, radioactive or magnetic 
spots and shapes for actuating the mechanisms. 

Some of the Why of using punched cards in scientific information work 
has been summarized in many publications. Hill, Casey and Perry,* in an 
article entitled, “Research and Chemical Information,” say: “There is 

* Hill, Norman C., Casey, Robert S., and Perry, James W., Chem. Eng. News, 25, 
970 (1947). 











































INTRODUCTION 


7 


grave danger that libraries of chemical information may become mere ware¬ 
houses of sheeted cellulose. The rapid and accelerating rate of increase in 
published scientific material can only render the situation more acute. 

“The chemical industry is faced with much the same problem as the 
telephone companies when they realized years ago that the time was fast 
approaching when there would not be enough qualified switchboard opera¬ 
tors to service the increasing number of calls. Just as the telephone industry 
developed the dial system, the chemical information field must develop 
newer, faster means and mechanisms for locating and correlating chemical 
information.” The same reasoning might be applied to fields other than 
chemistry. 

Frear, 4 in “Punch Cards in Correlation Studies,” says: “Experience gained 
in two years of work in this laboratory has led to the conclusion that 
punched cards are adaptable to a wide variety of chemical problems, par¬ 
ticularly those which deal with large groups of data, such as are encountered 
in surveys, correlations or extended experimental studies. 

“The particular problem we had here was a statistical investigation of 
the correlation between chemical structure and toxicity towards insects 
and fungi. From a search of the literature and other sources, data were col¬ 
lected on approximately 8,000 compounds, on each of which one or more 
toxicity tests had been made. With such a large number of compounds, 
and such a wide variety of constituent groups, counting and correlating 
the data by inspection promised to be a formidable task.” 

He then briefly describes his work in using hand-sorted cards and adds: 
“By slight modifications of these basic principles, it is possible to make cor¬ 
relation studies between chemical constitution and any desired property, 
chemical or physical.” 

Cox, Bailey and Casey 4 reported their work with hand-sorted cards in 
“Punch Cards for a Chemical Bibliography”: “In most of the bibliographic 
files which the authors have seen, the emphasis was on the manner in which 
the data are to be put into the file. The basis of the system described below 
is facility in getting desired data out of the file.” They also state, “Crane 
and Patterson in a discussion of the preparation of bibliographies have said: 
‘As to arrangement, there are at least four possibilities: by dates, by au¬ 
thors, by sources, and by subjects.’ With the proposed (punched-card) 
system, it is not necessary to choose one of the above categories for the 
arrangement of the references. One can arrange or segregate the cards 
according to any of those categories, and include other classes as needed, 
still using only one card for each reference.” 

4 Frear, Donald E. H., Chem. Eng. News, 23, 2077 (1945). 

' Cox, Gerald J., Bailey, C. F., and Casey, Robert S., Chem. Eng. News, 23, 1623 
(1945). 



Table 1-1 -Literature Searching Machines and Systems 


8 


PUNCHED CARDS 


Delivery 

Actual card containing ab¬ 
stracts, indexes, and bibli¬ 
ographic information. 

Document number is identi¬ 
fied which then serves as 
entree to document file ar¬ 
ranged in accession num¬ 
ber order. 

Document number is identi¬ 
fied which serves as entree 
to document file. Limited 
amount of bibliographic 
material or data may be 
printed on face of card or 
recorded on microfilm in¬ 
sert 

Storage 

Number of Documents 

Limited by tolerance of 
user to needling opera¬ 
tion* 

Limited in convenient 
use by quantity of 
document numbers 
that can be recorded 
on a single "aspect 
card" 

(a) Limited by toler¬ 
ance of user to sort¬ 
ing operations in 
fixed fields* 

(b) More convenient 

than (a) * 

(c) More convenient 

than (a) and (b) 
for certain applica¬ 
tions* 

Indexing Possibilities 

Direct coding: limited by 
number of holes in cards 
(up to about 200) 
Superimposed coding: lim¬ 
ited in number of index 
entires (2 to about 20) 
Relationships: not conven¬ 
ient to record complex 
relationships among in¬ 
dex entries 

Limited by increasing 
number of false combi¬ 
nations as "depth of in¬ 
dexing" increases 

(a) Similar to hand-sorted 
punched cards, above; 
certain relationships 
may be shown 

(b) Similar to hand-sorted 
punched cards, above; 
certain relationships 
may be recorded 

(c) Unlimited number of 
index entries per docu¬ 
ment; certain rela¬ 
tionships among index 
entries may be re¬ 
corded 

Manipulative 

Desired cards selected by 
manual manipulation of 
needles in holes or slots. 
Multiplicity of needles 
may be used with cer¬ 
tain auxiliary devices. 

(a) Columns of numbers 

on one "aspect" card 
matched visually 

against numbers on 
another aspect card 

(b) "Aspect" cards super¬ 
imposed, and desired 
document numbers 
detected visually by 
light passing through 
coincident holes 

Punched cards main¬ 
tained in drawers and 
fed into machine in 
stacks of up to 600. 

Examples 

^•3 
*? § 

« £ 

f! 

if 

a 8 

(a) Uniterm cards 

(b) Batten cards, Uniterm 
cards 

1* 5 S'; 

s |i| Jap 
I'JipUI 

<2g 11SS as 

S W W 

Name of Systems or 

Equipment 

i! 

l| 

* 

2. Hand-manipulated 
aspect cards 

(a) number match¬ 
ing 

(b) identification of 

pattern coinci¬ 

dence 

3. Machine-sorted 
punched cards 

(a) Fixed field 

(b) Intermediate 

(c) "Free field" 



INTRODUCTION 


9 


3 

a 

d 

d 

e 

if 

J 8 
3 I 

34 

II 


g 'S 

JS 

a t 

i i 

O ^ 

a « 

i 1 


3 

1-3 

1 § 
cd *» 

1| 

a 5 

I| 

<a 


s' 8 

■8|^_ 
u*i*i 
flslif 

si * 1 a 
- I 6 ! 5 
ag Si 

§ 11« 
Side 


c2 


8 a -S .2 I 


i § . 

J ! 

a J-3 9 

S' 8 j| 

i I g-s 


3 1 


O cu 


ss-ra 

| ®w o| 

a 5 * £ 



a o 
o 


f.. 

1 

5 1 


■O *2 13 -g 

5 « fi £ 

| 3 & 1 


iitli 

{K]|t 

I .J e 9 

a .s 1 v •« 
•a i » ° -S 

a M s s 2 

s 2 a •- S 
a *3 E a « 
£ 


o i g i * 

g si-ijil 

I || = W P §, 
E -i I'e I 5 5 
*JJsf 

I 2-tZ?2 1 § 



3 k *a s # 
8 (2 & 8 s 


a a s s ■»* 

4 § 11 s 
81 ^ 11 
| 1 ga* 
?•§ § 8 

iii'i 

lllah 

I'iisil 


> 

AS 


M I* 

j' II 
5 s 


■S ^3 

4*81 
C - 


I s I o 

a -a •” e _ . 

73 © ft >» « 2 .'2 ^ r’ 
3 5 fi w O 3 3 '3 


^ 3 © ”S 


>> « 


do-S'^H a 2 £ .2 a§ 


4 s 

n 


\ e •§ 

2 * 


r 


8 £ § 'S 


3 $ 


1! * si 
il}h 

ili'i 


IS 




.9 « 

a 
2 


W 


111 


*a$~ 

§ £ s 

5 111 

* £'il 

6 9 i*■* 

•s a -a s -J 

asall 

'flB .d 'd 


S 


6 


ill 

II! 

ill 

a js >5 


1 

8 


It. 


§ I 

► & a 

•sf a 

J3-g 

<8 SJ3 

|i 


if 

a S 


* £> 


B 

2 


|j!ll 

slsss 


•a 

S3 





10 


PUNCHED CARDS 


In “Some Applications of Punched-Card Methods in Research Problems 
in Chemical Physics,” King* summarizes his discussion of machine-sorted 
cards as follows: “The use of punched cards in research problems can be 
divided into three types: (1) large-scale repetition of simple operations, 
such as addition, subtraction and multiplication, modified if necessary by 
elaborate classification and selection; (2) a feasible stochastic or trial-and- 
error approach to the solution of problems; and (3) the construction of a 
representative sample of a population for statistical analysis. 

“These principles are illustrated by their use in preparing tables of ther¬ 
modynamic functions of compounds and spectrum analysis, and the cal¬ 
culation of the configuration entropy of high polymers.” 

Thus, the fundamental reason for using punched cards is that their use 
facilitates many routine and repetitive operations involved in the solution 
of certain intellectual problems. This is particularly true of problems in 
which large masses of data are involved. The machines can do some things 
which, due to their complexity and the amount of labor involved, could 
hardly be undertaken otherwise. 

Each new milestone in human progress presents new problems as well as 
opportunities. For example, the widespread use of printing as a means for 
recording and disseminating scientific information has been most bene¬ 
ficial. However, as Vannevar Bush 7 has pointed out, “Mendel’s concept 
of the laws of genetics was lost to the world for a generation because his 
publication did not reach the few who were capable of grasping and ex¬ 
tending it; and this sort of catastrophe is undoubtedly being repeated all 
about us, as truly significant attainments become lost in the mass of the 
inconsequential. The difficulty,” says Dr. Bush, “seems to me, not so much 
that we publish unduly in view of the extent and variety of present-day 
interests, but rather that publication has been extended far beyond our 
present ability to make real use of the record.” 

The punched-card technique is opening up new possibilities for coping 
with the growing mountain of research publication. These possibilities, 
however, are accompanied by problems, some of which are immediately 
apparent. One of the principal problems is the need to provide more precise 
methods for analysis and organization of information. 

Punched cards have aroused the interest of index and classification ex¬ 
perts, whose activities have of necessity been limited by their tools, namely, 
a set of pigeon holes or its equivalent for classification, and alphabetized 
lists of words either on bound sheets or in conventional card files for index¬ 
ing. The introduction of the punched-card technique has broadened the 
horizons of indexing and classifying and has opened new territory which is 
now being cultivated. 

• King, Gilbert W., J. Chem. Ed., 24, 61 (1947). 

7 Bush, V., Atlantic Monthly, 176, 101-8 (1945). 



INTRODUCTION 


11 


The extent to which the field has been cultivated is evident from the 
many publications discussing research, development, and applications of 
new devices, tools and systems which are recorded in the Bibliography 
(Chapter 30). 

The newer tools and systems for literature searching have been tenta¬ 
tively classified recently by Kent and Geer 8 in Table 1-1. 

Punched cards and related devices cannot make all complex problems 
simple, but they can make some complex problems less complex and some 
simple problems less tedious and time-consuming. 

* Allen Kent and Harriet Geer, “Searching the Chemical Literature Mechani¬ 
cally,” Paper presented before the American Chemical Society, Sept. 1966, Atlantic 
City, N. J. Table reprinted from A. Kent, Am. Doc., 8, No. 2, 150-151 (1957). 



Chapter 2 

ELEMENTARY MANIPULATIONS OF 
HAND-SORTED PUNCHED CARDS 


Robert S. Casey 

W. A. Sheaffer Pen Co., Fort Madison, Iowa 
AND 

James W. Perry 

Center for Documentation and Communication Research, Western 
Reserve University, Cleveland, Ohio 

The How of using punched cards has two aspects. One of these is mechan¬ 
ical and involves learning how to punch, sort and otherwise manipulate 
the cards. The other aspect is intellectual and concerns the necessity of 
analyzing the subject matter to which punched cards are to be applied. 
One must determine what meanings are to be assigned to the holes in the 
cards and what mechanical manipulation of the data is required, as dis¬ 
cussed briefly under “Subject Analysis and Coding” later in this chapter. 
Although it is not possible to discuss either of these aspects apart from 
the other, this chapter is concerned principally with the first aspect—the 
basic mechanics of punched-card techniques. 

Description of Cards 

Hand-sorted punched cards may be obtained commercially in a variety 
of sizes, from x up to 8 x 10^ inches. One widely used type 
(“Keysort”) has one or two rows of 3-6-inch holes on M -inch centers paral¬ 
lel to the edges of the cards, with the first row 3^6 inch from the edge. An¬ 
other type (“E-Z Sort”) has elliptical holes spaced 6 to the inch. With all 
types of hand-sorted, edge-punched cards, the holes along the edges occupy 
only a small fraction of the total card area. Consequently, most of the 
area on both sides of the card is available for writing, typing or printing 
references, abstracts, observations, and numerical data, as well as for 
coding directions and attaching pictures, clippings and other thin, flat 
material. 

One corner of each card is cut off so that it is possible to see at a glance 
that all the cards are right side up and facing the same way. The holes in 
the other three corners are never punched. They are used to arrange the 
cards right side up and to see that they face forward should they become 
mixed (see page 18). 


12 



MANIPULATIONS OF HAND-SORTED PUNCHED CARDS 


13 


BASIC SORTING OPERATION 
Direct Sort of Outer Row Holes 

The basic sorting operation separates the cards punched to form a notch 
in a given position from those not so punched. As shown in Figure 2-1, in¬ 
sertion of the tumbler or sorting needle into the hole in question, and then 
raising the tumbler or needle, permits the cards punched (notched) in that 
position to fall, unless they are prevented from doing so by the friction of 
other cards. In order to facilitate the dropping of cards punched in the 
position being sorted, the following technique has been developed. 

Figure 2-1. Remove from the file a group of cards not more than 2 inches 
thick and place them in a vertical position on the alignment block with the 
hole portion to be sorted at the top. Jog the cards against the vertical edge 
of the alignment block to align the holes. Support the cards with the left 
hand. Place the left thumb adjacent to the hole to be sorted and compress 
the cards with the thumb and fingers of the left hand. Grasp the handle of 
the tumbler firmly in the right hand, palm underneath the handle. Keep 
the tumbler horizontal at all times to prevent the cards falling off the needle 
or sliding back against the handle. Insert the needle into the hole to be 
sorted, guiding it with the left thumb. Push the needle through the cards, 
leaving at least one inch between the tumbler handle and the front card. 

Note —The alignment block, shown under the cards in Figures 2-1 to 2-6 
inclusive, is a sheet metal device which fits against the front edge of the 
desk. A portion at the right edge is bent into a vertical position perpendicu- 



Figure 2-1. Start of single needle direct sort. 


14 


PUNCHED CARDS 



Figure 2-2. Second step in sorting cards. 


lar to the front edge of the desk. This device is not absolutely necessary, 
but assists in sorting the cards as described below. 

Figure 2-2. Support the cards loosely with the left hand at the left 
vertical edge of the cards. Swing the handle of the tumbler toward the 
left, pushing the cards toward the right with your left hand. This cramps 
the cards in a diagonal fashion as shown in Figure 2-2. While the cards are 
in this position, grasp them firmly between the thumb and fingers of the 
left hand. 

Figure 2-3. While holding the left edge of the cards firmly with the 
left hand, move the handle of the tumbler toward the right, back to its 
original position. This causes the cards to fan out and separate as shown 
in the illustration. Hold the cards in this position and with both hands 
lift them several inches above the alignment block. Keep tumbler horizontal. 

Figure 2-4. Release the left-hand grip on the cards, at the same time 
giving them a slight downward jerk with both hands. Hold the left hand 
as shown in the diagram, forming a U, and lightly support the cards which 
drop. Do not grasp falling cards tightly. If the cards are compressed at this 
point, it will hinder their separation. Swing the sorting needle back and 
forth gently from left to right so that the suspended cards pivot about the 
needle. This motion facilitates dropping the cards being sorted. 





Figure 2-3. Third step — cards are fanned out. 




















16 


PUNCHED CARDS 



Figure 2-5. Fifth step—vertical edge on alignment block helps separate dropped 
cards. 


Figures 2-5 and 2-6. Move the tumbler with the remaining cards 
hanging on it toward the right so that the lower edges of the cards just 
clear the vertical portion of the alignment block, which then retains the 
dropped cards. With the left hand place the dropped cards on the desk at 
the left of the alignment block. The right hand now holds the tumbler 
with the rejected cards still suspended. Place the rejected cards on the 
alignment block before removing them from the tumbler. Twist the tumbler 
in a vertical plane, moving the handle downward. This spreads the cards 
so each one is higher than the one in front of it. Scan the top of the cards 
just sorted to see if any cards edge-punched in that hole failed to drop. 
If so, remove them and place them with the selected cards. 

The series of operations just described can be completed in about 15 
seconds or less by a person who has had only a moderate amount of ex¬ 
perience. The operations are then repeated on additional groups of cards 
until the complete file has been sorted. 

If the position to be sorted is near the left edge of the card, there may 
not be enough space between the tumbler and the left hand to spread the 
cards as illustrated in Figure 2-3. In this case, move the left hand to the 
bottom of the cards directly below the tumbler and twist the tumbler in a 



MANIPULATIONS OF HAND-SORTED PUNCHED CARDS 


17 



Figure 2-6. Sixth step—separation of selected from rejected cards is completed. 


/ 

vertical plane, moving the handle down. Grasp the cards at the bottom 
edge firmly between the thumb and fingers of the left hand and twist the 
tumbler back to its original horizontal position. This gives the same result 
as shown in Figure 2-3. 

Sorting Operations Involving Double-Row Holes 

Double-row holes may be punched in three different ways as shown in 
Figure 2-7. Cards deep-punched in a given position may be separated 
from all the others by inserting the needle in the inner hole and carrying 
out the sorting operation described for sorting outer row holes. Similarly, 



t t t 

1 SHALLOW 
1 INTERMEDIATE 
DEEP 

Figure 2-7. Three different ways of punching each position in a double row of holes. 






18 


PUNCHED CARDS 


separating deep-punched and shallow-punched cards from cards not 
punched at all, or only intermediate-punched, is exactly the same as sort¬ 
ing single-row holes. 

Separation of intermediate-punched cards from the others is effected 
by inserting the needle in the inner hole, spreading and dropping the cards 
as already described. As a result, the intermediate-punched cards will drop 
about 34-inch, but will not fall off the needle. Keeping the tumbler hori¬ 
zontal, but not allowing the cards to rest on the table, jog the cards against 
the vertical portion of the alignment block to bring them into horizontal 
alignment. Now, grasp the cards firmly between the thumb and fingers 
of the left hand. Withdraw the tumbler and insert it into one of the comer 
holes which are never punched (notched). Again fan out the cards and the 
intermediate-punched ones will now fall clear. 

Arranging the File for Sorting 

It has been assumed in the discussion up to this point that all the cards 
in the file have been right side up and facing forward. If the cards are not 
so arranged, this will be evident from glancing at the corner cut-off. The 
holes in the other three corners, which are never punched, are used to 
arrange the file properly for sorting. If the upper right hand corner of the 
file is clipped, insert the tumbler into the hole showing at the upper right 
corner of the file, and proceed with the sorting operation as previously 
described. The cards which drop are right side up and facing forward. 
Repeat, after turning the cards which remain on the tumbler through 180° 
in their own plane. Then rotate the cards which still remain on the tumbler 
180° perpendicular to their plane, and again needle the upper right hole. 

ELEMENTARY SUBJECT ANALYSIS AND CODING 

Specific meanings must be assigned to the holes with consideration for 
whatever subsequent sorting may be required. In general, this is done with 
one or both of two purposes in mind, namely, selecting certain cards from 
the file or arranging all the cards in a given order. 

Coding is matching the idea, datum or concept with a punched or notched 
hole or a pattern of punches. For convenience, a set of symbols is inter¬ 
posed between concept and card. For example, the punching or notching 
positions on cards are usually numbered. Then the file is “coded” by mak¬ 
ing a list of subject headings and assigning one of the numbers or combina¬ 
tions of numbers or other symbols indicated on the card to each entry. 
The term “code” is also used to refer to the pattern of symbols on the 
card. Letters of the alphabet can be indicated by directly labeled positions 
or by numbers. Confusion may arise because on different occasions a 
letter or a number may be either a concept or a symbol. 



MANIPULATIONS OF HAND-SORTED PUNCHED CARDS 


19 


Remember that the symbols are merely an intermediate convenience. 
The code is: a certain pattern of punches equals a certain concept. 

In setting up a punched-card file one must analyze the purposes of the 
file and the ways in which the data on the cards are to be used. Thus, with 
a bibliographic file, one should be able to select the cards bearing references 
concerning any one of the subjects of interest to the bibliographer. Also, 
it may be desirable to arrange the whole file alphabetically by authors or 
chronologically by date of publication. If one wishes also to select cards 
according to author and date of publication, one must choose the appro¬ 
priate type of code. 

Direct Coding. In this, the simplest form of coding, a separate mean¬ 
ing is assigned to each hole. All the cards on which the meaning is coded 
(e.g., all references published in the nineteenth century, or those describing 
analytical procedures or chemical compounds which contain the hydroxyl 
group) are separated with a single pass of the sorting tumbler. The appli¬ 
cation of direct coding has its limitations. For example: There may not be 
enough holes on the card to code all of the desired data and more passes 
of the sorting tumbler are required for serial sorting than are required for 
numerical codes. 

Numerical Codes. A numerical code might be called a “combination” 
code since one or more holes may be punched to represent a single number, 
letter or other entity. The most commonly used numerical code is illus¬ 
trated in Figure 2-8. 

By punching various combinations of the four holes marked, respec¬ 
tively, 7, 4, 2, and 1, one may code any number from zero (no punching) 
up to and including fourteen (all holes punched). Such a group of holes 
is called a “field.” This code is a modification of the 1, 2, 4, 8, 16 ... series; 
7 is used instead of 8, so that with four positions any digit may be indicated 
by punching not more than two holes. By using one such field each for 
units, tens, hundreds, etc., relatively few holes are required to code any one 
large number that may be desired (Figure 2-9). 

A file of cards coded in this way is arranged into numerical order by 
sorting in order from right to left each hole in the numerical fields and 
placing the cards which drop at the back of the pack before sorting the 
next hole. Detailed instructions for this sorting procedure are given later 
in this chapter. 


oooo 

7 4 4 1 

X 


M 


Figure 2-8. A numerical sorting field. 



20 


PUNCHED CARDS 


■ 







NO. t2679 



Figure 2-9. A 5-Digit numerical sorting field. 



Figure 2-10. A 2-field, 5-hole alphabetic code. 
Letters coded: A R. 


Although large numbers can be coded in numerical fields, only one num¬ 
ber can be coded in any one field on any one card. Thus, such coding is 
used for numbered lists of data which are mutually exclusive such as serial 
numbers. Although such coding provides a convenient means for sorting 
the file into serial order, unequivocal selection of a card coded for a given 
number is not possible. (See Selector and Superimposed codes, below) 

Alphabetical Codes. Alphabetical codes, like numerical codes, are 
based on the use of combinations of holes, with the exception that the 
coding represents letters of the alphabet instead of numbers. 

A commonly used alphabetical code which is, in fact, a variation of the 
numerical code, is illustrated in Figure 2-10. The letters A to M, inclusive, 
are numbered consecutively 1 to 13, and the appropriate number is coded 
to represent the desired letter. If the desired letter is in the second half of 
the alphabet, the letters N to Z are coded in the same manner, with the 
additional punching of the N-Z hole. 

Alphabetical codes, as well as numerical codes, can be used to sort a 
file into serial order; in fact, the same general comments made concerning 
numerical codes also apply to alphabetical codes. 

Chronological codes can be made up by adapting a direct or combination 
code to the need of the user. If it is necessary to code only the year, one 
may use two 7, 4, 2, 1 fields for the units and decades, respectively, and 
one or more other holes for the century, depending on how long an inter¬ 
val is to be covered. 

Selector Codes. In some punched-card files it is desirable to be able 
to select all cards of a certain category. If there are more categories than 
holes available for direct coding, selector codes are necessary. Selector 
codes are special combination codes so conceived that, in the simplest 
case, two holes, no more and no less, are punched in each field to represent 
each symbol. Then, when two sorting needles are inserted into the holes 
representing the desired symbol and lifted, only the desired cards drop. 











MANIPULATIONS OF HAND-SORTED PUNCHED CARDS 


21 



Figure 2-12. A 5-hole triangle or pyramid selective code, with symbols arranged in 
proper order to permit serial sorting also. 1 Number coded 3. See also Fig. 21-1. 


A commonly used selector code has the positions 0 (zero) and SF (single 
figure), in addition to 7, 4, 2, 1. To code 1, 2, 4, or 7 the SF hole is punched 
in addition to the numbered hole. The other digits require the punching 
of two of the numbered holes. To code zero the 0 position alone is punched 
(Figure 2-11). 

Another simple selector code is the triangle or pyramid type shown in 
Figure 2-12. It is coded by punching the holes at the tops of the mutually 
perpendicular diagonal columns which intersect at the desired symbol. In 
the example shown, the two columns are those containing the numbers 0, 
1,3,6 and 5, 4, 3, respectively. 

These selector codes permit only one number to be coded in common 
with the other combination codes described previously. 

Double-Row Coding. The codes described above are for a single row 
of holes. A double row of holes increases the coding possibilities in a given 
amount of card space since each position in a double row may be punched 
in any one of three ways, using the shallow, intermediate, or deep punch, 
(Figure 2-7). The inner and outer rows of holes can also be coded and sorted 
independently, by punching the outer row of holes with the shallow punch, 
and the inner row with the intermediate punch, which clips open the space 
between the inner and outer hole. 

Superimposed Coding. The direct and combination coding schemes 
already described are often inadequate to code the number of subject con¬ 
cepts that are required. Superimposed coding permits one to code on each 
card several concepts selected from a list of a great many more concepts 
than there are holes in the card. Each concept is coded by notching two or 
more of the positions along one edge of the card. 

1 Cox, Gerald J., Robert S. Casey and C. F. Bailey, J. Chem. Ed., 24, 65 (1947). 




22 


PUNCHED CARDS 


The system can be designed with the combinations assigned randomly 
so that the chance of getting unwanted cards with a given sort is negligible. 
For example, if one sorts for “disease,” which is assigned holes numbered 
7 and 13, and for “antibiotic,” 4 and 27, and for “animal,” 15 and 19, one 
would get only a negligible number of “extra” cards coded 4 and 13, 7 
and 19, 15, and 27. 

Various modifications and applications of the principle of superimposed 
coding are described in Chapters 10, 15, 18, 21 and 23, and in various refer¬ 
ences in the bibliography. 

Subject Analysis. Some sort of analysis of subject matter is necessary 
no matter what kind of file or system is to be set up. A collection of re¬ 
prints or a file of references on plain cards may be filed in alphabetical 
order by author, grouped according to subdivisions of the main subject, 
or indexed in a notebook. Often additional cross-reference cards are made 
to cover different subject aspects of a given reference. Books and pamphlets 
are placed on shelves in some sort of order. 

When setting up a punched-card file, however, special consideration 
must be given to subject analysis in order to make full use of the advan¬ 
tages offered by punched cards (Chapter 1). New considerations are neces¬ 
sary, or rather a new viewpoint concerning the extensions and combina¬ 
tions of older considerations is necessary. 

One of the most valuable properties of punched cards is their multi¬ 
dimensional or multi-aspect coding possibilities. Each of various independ¬ 
ent aspects of the subject matter can be coded independently. For example, 
substances can be listed according to their composition or form, together 
with their chemical or physical properties or their functions such as “sol¬ 
vent,” “antioxidant,” “lubricant,” and “fungicide.” Another aspect is 
“processes and procedures,” such as “analyze” or “measure,” “oxidize,” 
“distill.” Others are “conditions,” “energy manifestations,” and “struc¬ 
tures” such as “machines” and “apparatus.” 

Then several broad entries from various aspects can be combined to 
define a specific bit of information. A thermometer is “apparatus” to 
“measure” “thermal condition.” One can select cards bearing information 
about “antibiotics” in the “treatment” or “therapy” of “gastrointestinal” 
“diseases” in “animals.” 

Avoid making the entries under each aspect too specific. Make the entries 
as broad and generic as is consistent with the subject matter in the file. 
The coding should be done only after careful study of the references. It 
may be necessary to make a generic search, say for “antibiotics.” If anti¬ 
biotics have been coded individually, sorting for “penicillin,” “strepto¬ 
mycin,” and the others one by one would be required. Furthermore, if one 
is seeking references on an individual antibiotic, it may be easier to go 
through several dozen cards coded “antibiotics” by hand, and pick out 



MANIPULATIONS OF HAND-SORTED PUNCHED CARDS 


23 


the relevant ones than it would be to make a more elaborate and specific 
sort. 

Don’t try to set up a classification scheme or schedule of subject entries 
a priori. Study the references. Think of the ideas which are discussed, not 
the words used to describe the ideas. Ask yourself, “What are some of the 
questions this document would answer?” “Why would I ever want to find 
this document?” 

Think generically. Don’t code individually “children,” “men,” “women,” 
“Caucasians,” “Mongolian idiots” when you can use “humans” and not 
cover too many references. Start with “humans,” and leave some of the 
holes in the cards unassigned so you can subdivide later, if necessary. 
For some files “humans” might be too specific, “animals” or “living or¬ 
ganisms” could be used. 

Each subject entry should cover a carefully defined area of meaning. 
The entry does not need to be a single word, it can be a phrase, a sentence, 
a paragraph, or a diagram of a structure. The entry can be any notion the 
human mind may conceive, but its area of meaning should be carefully 
defined. 

Subject analysis and coding are discussed in more detail in Chapters 
18, 24, and 25 and in most of the chapters in Part II. Coding of chemical 
compounds is described in Chapter 22. The mathematical analysis of 
coding is developed in Chapter 21. 

SPECIAL SORTING TECHNIQUES 

As already pointed out in discussing “Basic Sorting Operations,” it is a 
very simple process to separate cards punched (notched) in a given posi¬ 
tion from those not so punched. By combining a succession of simple sort¬ 
ing operations with certain special codes, it is possible to effect the arrange¬ 
ment of a file into a numerical or alphabetic sequence. The technique of 
doing this is duscussed in the following paragraphs. 

Sequence Sorting. Perform the sorting operations illustrated in Figures 
2-1 through 2-6 on each hole and in order from right to left through the 
fields composing the numerical or alphabetical code. The cards which drop 
after each sort are placed in back of the cards which remain on the tumbler 
before the next hole is sorted. All the cards which drop must be kept in 
their relative order. All cards which fall out of position, or fail to drop 
from the group hanging on the tumbler, must be laid aside and filed manu¬ 
ally after the sorting is completed. 

After each sort, jog the dropped cards into alignment with the left hand. 
Then, with the right hand still grasping the tumbler handle, return the 
rejected cards to the alignment block in front of the dropped cards. Glance 
along the edges of the cards remaining on the tumbler at the hole just 
sorted to see if any cards failed to drop. Next bring the dropped and re- 



24 


PUNCHED CARDS 


jected cards into alignment, allowing the needle to slip into the groove 
in the group of cards just dropped. Grasp the cards firmly between the 
thumb and fingers of the left hand, with the thumb adjacent to the left 
edge of the next hole to be sorted. Remove the tumbler from the hole just 
sorted and insert it into the next hole to the left, repeating the sorting 
operation just described. After sorting the last hole at the left end of the 
coded portion, the cards will be in serial order. The above process is called 
fine sorting. 

If there are more cards to be sequence sorted than can be handled on 
the tumbler at one time, the above fine sorting procedure must be pre¬ 
ceded by a breakdown sort. In general, the breakdown sort is carried out 
by sorting each hole from left to right and by placing in separate piles the 
cards which drop when each hole is sorted until the whole file is arranged 
in piles small enough to be fine sorted. The following example will illustrate 
the procedure: 

9,999 cards, numbered from 1 to 9,999 and coded by the 
7, 4, 2, 1 system, are to be arranged in numerical order. 

Take from the file a group of cards not over two inches thick. 

(1) Sort through the 7 hole in the “thousands” field. 

7xxx,* 8xxx, 9xxx cards drop and are placed at the left rear 
portion of the desk as the first pile. 

(2) Sort the 4 hole in the “thousands” field. 4xxx, 5xxx, 

6xxx cards drop and are placed in the second pile at the right 
of the first pile. 

(3) Sort the 2 hole in the “thousands” field. 2xxx, 3xxx 
cards drop and are placed in the third pile at the right of the 
second. 

(4) Sort the 1 hole in the “thousands” .field, lxxx cards 
drop and are placed in the fourth pile at the right of the third. 

Cards numbered 1 to 999 remain on the needle and are placed 
in the fifth pile at the right of the fourth. 

(5) Repeat operations (1) to (4) with successive groups of 
cards until all cards have been distributed to the five piles. 

(6) Repeat operations (1) to (4) in the “hundreds” field 
on the fifth pile, which contains cards numbered under 1000. 

This makes additional piles as follows: 

7xx-8xx-9xx 

4xx-5xx-6xx 

2xx-3xx 

lxx 

1 to 99 

* “x” indicates any digit, 0 to 9. 



MANIPULATIONS OF HAND-SORTED PUNCHED CARDS 


25 


(7) Fine sort the 1 to 99 group and place it in the front of 
the file storage drawer. Fine sort the lxx group and place it 
in the file behind the 1 to 99 group. As each group is fine 
sorted, put it in the file in this order. 

(8) Sort the 1 hole in the “hundreds” field in the 2xx-3xx 
group. Fine sort the 2xx group, which remains on the needle, 
and place it in the file. Fine sort the 3xx cards, which have 
dropped, and place them in the file. 

(9) Sort the 2 hole and then the 1 hole in the “hundreds” 
field in the 4xx-5xx-6xx group. Fine sort each of the three 
resulting piles in this order: the 4xx cards, which remained 
on the needle; the 5xx cards, which dropped when the 1 hole 
was sorted; and the 6xx cards, which dropped when the 2 hole 
was sorted. Then place them in the file. 

(10) Sort the 2 hole and the 1 hole in the “hundreds” field 
in the 7xx-8xx-9xx group. In the same order as described 
under (9) fine sort the three resulting piles and place them 
in the file. 

(11) Repeat operations (6) to (10) with the fourth pile, 
which contains cards numbered 1000 to 1999. 

(12) In their turn, sort the other piles, first into thousands 
and then into hundreds. Then fine sort and file each hundred 
as outlined in the preceding steps. 

If two persons are sorting, the breakdown sorting should be done by one 
person and the fine sorting by the other. 

Multiple Sequence Sorting. A punched-card file may be arranged in 
serial order according to one category (“Major” item), and, at the same 
time, in serial order according to a second category (“Minor” item), under 
each item of the first. 

Multiple sequence sorting is used in correlation studies, such as the 
physiological properties versus-, chemical structure discussed by Frear in 
Chapter 22. It may also be used for arranging a bibliographic file in chrono¬ 
logical order according to the date of publication of each subject covered 
by the file. To accomplish such an arrangement it is necessary to sequence 
sort the “Minor” items first, then sequence sort the file according to the 
“Major” item. This is clarified by remembering that we are doing the 
same thing when we sort several fields of a numerical code into serial order. 
The “units” are arranged first in serial order, then the “tens,” and so on, 
until the “units” are in order under each digit of the “tens,” the “tens” 
are in order under each dipt of the “hundreds,” and so on through as many 
places as there are in the code. This allows more than two categories to 
be put into serial order, one within the other. 



26 


PI NCHED CARDS 


If there are more cards to be sorted than can be handled conveniently 
at one time, breakdown sort the “Major” item. When that is complete, 
fine sort according to the “Minor” item. This becomes clear if one con¬ 
siders the codes of the “Major” and various “Minor” items as one con¬ 
tinuous numerical code with each item, in order, occupying one decimal 
place. 

Sorting Selective Codes. For sorting selective codes additional needles 
(without handles) and a tumbler are required. The needles used may be 
either metal rods or metal or plastic knitting needles, 2 millimeters in 
diameter and about 10 inches long. 

In the selector codes two needles are inserted in the appropriate holes in 
each field (except for 0 in the Figure 2-11 code) to select cards punched 
for the desired number, letter or coded symbol. 

Insert the tumbler into one of the holes required (if possible, near the 
center of the edge of the card for balance). Insert the loose needles in the 
other required holes. 

Pivot the tumbler in the vertical plane, pressing the handle downward. 
Firmly grasp between the thumb and fingers of the left hand the bottom 
edge of the cards directly below the tumbler. Pivot the tumbler back to 
horizontal. Raise the cards a few inches and release the left hand, dropping 
the punched (notched) cards as described under “Direct Sort of Outer Row 
of Holes.” This process is repeated with additional groups of cards, until 
the whole file has been sorted. 

A group of cards not more than 1^ inches thick should be taken for 
each selective sort. However, the number of cards which can be sorted 
conveniently at one time will be determined by experience. 

This method of selective sorting requires that the needles be reinserted 
individually in each group of cards sorted. When the punched-card file is 
large and considerable selecting is to be done, it may be advantageous to 
use a selector unit described in Chapter 3. This unit has a metal bar pro¬ 
vided with openings spaced the same as the holes along the edge of the 
card. The ends of the sorting needles are inserted into those openings which 
will space the needles suitably for the sort being contemplated. Thus, any 
number of groups of cards may be sorted without resetting the individual 
needles. 

Most selector codes may also be serially sorted. The code illustrated in 
Figure 2-11 can be serially sorted by ignoring the 0 and SF positions and 
sorting the 7-4-2-1 fields exactly as described above for numerical codes. 

The code illustrated in Figure 2-12 can be fine sorted serially by sorting 
each hole in turn from right to left as described previously for the 7—4—2—1 
code, except that it is not necessary to sort the hole at the right. The fine 
sort can be started at the second hole from the right. The breakdown sort 



MANIPULATIONS OF HAND-SORTED PUNCHED CARDS 


27 


is performed in a similar manner to that described for numerical codes. 
The holes are sorted from left to right, and the cards which drop as each 
hole is sorted accumulate into separate piles. Superimposed codes are 
sorted the same as selector codes. 

SUPPLEMENTARY NOTES ON HAND-SORTED CARDS 

Punched cards may be obtained with special forms printed on the card 
and the coding printed adjacent to the holes. Manufacturers also have 
certain standard cards which are more economical than specially printed 
ones. When standard cards are used, the coding and other matter may be 
overprinted or multigraphed, or the cards may be punched and sorted by 
referring to a master code card. 

If the punching is to be done with a hand punch, the holes to be punched 
should be marked first. Avoid using soft crayon for this purpose because 
pieces of the crayon scrape off at the edges of the hole or card and cause 
unsightly streaks on the cards. 

Precaution must be taken to prevent the cards from sticking together. 
To this end, avoid permanently tacky adhesives when attaching material 
to the cards. Paper, illustrations and samples of thin material may be 
attached to the cards by non-tacky plastic-base or heat-sealing adhesives. 
Even the buckling caused by water-base adhesives may not prohibit their 
occasional use. Do not put rubber bands around the cards. Rubber ages 
rapidly under tension and will leave tacky spots on the cards. 

Bent or wrinkled cards will generally sort as well as, or better than, new 
cards. However, avoid adjacent cards becoming deformed in the same 
pattern. If the file becomes cramped and the corners or edges of the cards 
are bent so the cards tend to “nest,” sorting will be hindered. 

If these measures are heeded, a punched-card file may be stored and 
handled the same as an ordinary card file. 

The information given in this chapter should be sufficient to enable an 
individual to perform the punching and sorting operations required to set 
up and use a simple file of hand-sorted punched cards. 

An understanding of the mechanical features of punched cards is, how¬ 
ever, only the first step toward their successful use as tools for handling 
information. Applying punched cards successfully to a problem also re¬ 
quires careful analysis of the subject matter involved, and its segregation 
and use. The chapters mentioned at the end of the elementary “Subject 
Analysis” section above, should be consulted. 

It should be pointed out, however, that long and detailed study is not 
necessary before a person can obtain good results with a punched-card 
file. The beginner should strive for simplicity in coding. He should avoid 
complicated codes and too fine a breakdown of subject matter. It should 



28 


PUNCHED CARDS 


be remembered that it takes only a few minutes to glance over the mate¬ 
rial entered on several dozen cards. For this reason a code can be regarded 
as quite satisfactory if it permits sorting operations to select all the desired 
cards, even though they may be accompanied by a moderate number of 
undesired cards. 

It is advisable at the beginning not to attempt to set up the code in 
final form. It is usually better to start out with a skeleton code built up of 
broad topics whose usefulness for effecting sorting operations cannot be 
doubted. This approach makes it possible to keep many holes in reserve 
for such finer analysis of the subject matter as may prove necessary. 

Although a punched-card file can be kept in random order and sorted 
when necessary, it is often convenient to keep the file arranged in some 
order or segregated into broad subject categories. 

Also it is convenient to assign a serial number to each reprint or other 
document as it is acquired, and to code, or at least write the number on 
the punched card. The documents can be kept in serial order and the cards 
in alphabetical order by author. Then a reference can be located by author 
or by serial number. 

Edge-marking with tabs* and with pen or pencil*- 4 and edge-notching 6, *■ 7 
of plain cards have been suggested. The edges of a pack of plain cards can 
be fanned out and then scanned for a colored mark or notch in a given 
position on a given edge. This simple technique of broad subject searching 
might be tried if one has a file of plain cards, preliminary to adoption of 
punched cards. 

Suggestions have been made for making one’s own punched cards. A 
template and hand punch,® a drill press, 9 a specially constructed jig, 10 and 

* Reumuth, H., “The Indexing of Chemical Compounds. A Contribution to the 
Problem of Organization of the Literature,” Z. Angew. Chem., 41, 1204-7 (1928). 

* Thurstone, L. L., “The Edge-Marking Method of Analyzing Data,” J. Am. 
Statistical Assoc., 43, 451-62 (1948). 

4 Lester, A. M., “The Edge Marking of Statistical Cards,” J. Am. Statistical 
Assoc., 44, 293-4 (1949). 

* Aldrous, J. G., “Simple Method for Cross-Indexing a Reference File,” Science, 
106, 109 (1947). 

* Campbell, D. J., “The Use of Notches in Cards as a Means of Signalling Informa¬ 
tion,” J. Documentation, 9, 224-5 (1953). 

7 Schlink, F. J., “Getting the Most out of Index Cards,” Industrial Management, 
55, 135-8 (Feb. 1918). 

* Begun, George M., “Making Your Own Punched Cards,” J. Chem. Educ., 32, 
328 (1955). 

* Thomas, George R., “The Preparation of Punched Cards for Indexing Informa¬ 
tion,” J. Chem. Educ., 29,406 (1952). 

10 Thomas, Carl O., “A Jig for Preparing Edge Punched Cards,” J. Chem. Educ., 
34, 241 (1957). 



MANIPULATIONS OF HAND-SORTED PUNCHED CARDS 


29 


a punch for preparing papers for plastic binders, 11 have been used to per¬ 
forate the edges of either new cards or the cards in an existing file. 

By following the simple precepts suggested in this chapter the beginner 
will find it possible to use his own experience as a basis for improving his 
skill in using punched cards. 

11 Cullman, Ralph E., (Letter to the Editor), J . Chem. Educ 30, 246 (1953). 



Chapter 3 

COMMERCIALLY AVAILABLE EQUIPMENT 

AND SUPPLIES* 

Thomas H. Rees, Jr. 

Center for Documentation and Communication Research, Western 
Reserve University, Cleveland, Ohio 

This chapter is divided into five parts. The first deals with edge-notched 
punched cards and related equipment; the second with tabulating type 
punched cards; the third is concerned with supplements to punched cards; 
the fourth with other types of equipment related to information process¬ 
ing; and the fifth with ancillary equipment for punched-card systems. 

All of the material presented here has been reviewed by the manufac¬ 
turers or suppliers mentioned, and in most cases it includes descriptions 
taken from their literature. Replies from several commercial organizations 
were not received in time for inclusion and therefore descriptions of their 
products have been omitted. 

EDGE-NOTCHED PUNCHED CARDS 
The Keysort System 

Royal McBee Corporation 
Port Chester , N. Y. 

The Keysort systems employ the marginal hole punched-card principle 
described in Chapters 1 and 2. All punching and sorting machines are 
manually operated, and their operation can be learned in a few minutes. 
Reasonable operating speeds are approached in a few days and usually at¬ 
tained within a week. 

Royal McBee maintains representatives in more than 70 cities of the 
United States and Canada to service their systems and related equipment. 
Key punches, batch grooving machines, and card counters are available 
on a rental basis only, whereas all other equipment and supplies listed 
below may be purchased. No special wiring or building construction is 
necessary. 

* Submitted in partial fulfillment of the requirements for the degree of Master of 
Science in Library Science, Western Reserve University, Cleveland, Ohio. 


30 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 31 






1 7 4 1 2 1 1 

7 42 * 1 | 7 4 1 * I I 

1-ROM unn 

iiMM-1 cW«*wbiift. T -THHi- 

HiWM 

H- 1 - ZZR 




TVia KEYSORT CARD wm IMPRINTED mi KEYPUNCHED 
•iadtaasoolr by tW ROYAL McBEE DATA PUNCH 

J_I_I_I_L 


t.*.T!!S 

tv-* 


Ric:::.r.D roe 
35-517 


F 


211 


i| DATA PUNCH 1 

N*w yud .. 


1 cl KEYSORT 
.0* tt» joblnttfty 


Itllitu 


1 r t v ~ r | T H, iyp r "TfT 


• » 


r^Prrr,',; T7T—zti— ryr-T 


Figure 3-1. Keysort card punched and imprinted by Data Punch. 


Keysort cards (Figure 3-1) are manufactured in sizes ranging from 
1^2 x 2 Yi to 8 x 10^ inches. Larger cards are sometimes supplied but 
they are not recommended by the manufacturer. Various weights and 
grades of ledger and card stocks are available depending on the application. 
Widely used for bibliographies is a 50 per cent rag content stock 0.0085- 
inch thick, which requires a filing capacity of 1-inch linear drawer space 
for each 100 cards. All cards are perforated at the factory to provide either 
a single or double row of ^-inch holes adjacent to one or more edges. The 
holes are spaced four to the inch. With two parallel rows of holes there are 
8 coding positions available for each 1 inch of card edge space. Since it is 
usually not necessary to have more than %-inch around the edge of 
the card for coding purposes, the major portion of the card is available 
for written or typed information. The card stocks generally lend readily 
to duplication by any of the commonly used office processes. A Keysort 
card cut from multicopy index bristol stock makes a satisfactory hectograph 
master. Manufacturing accuracy precludes alignment difficulties. Either 
16- or 3o-mm microfilm frames can be inserted in the body of the card 
with no reduction in coding capacity and only a small reduction in writing 
space. Cards made from Ozalid translucent stock can also be supplied. 
Such cards permit copying onto Ozalid sensitized Keysort cards. 

Keysort card savers (Figure 3-2) are gummed strips of paper perforated 
to provide holes corresponding in size and spacing to those along the edges 
of the Keysort cards. As shown in Figure 3-3 the card savers may be stuck 
over the edge of the card to restore the holes where they have been punched 
in error, to change the coding wffiere it is desired or to repair the factory 
perforation where it may have been damaged. They may also be used as a 
fringe to join two cards along one edge. Thus, where information is too 
voluminous to be entered on both sides of a single card, it may be con¬ 
tinued onto another card. 









32 


PUNCHED CARDS 




Figure 3-4. Keysort hand punch. 


Keysort hand punches (Figure 3-4) are used for punching the cards. 
Three styles are available: shallow punch for single-row coding; deep 
punch for double-row coding (if card is inserted only halfway into throat 
in punch head, outside row is shallow-punched); and intermediate punch 
for inside double-row coding. (Intermediate punching is described in Chap¬ 
ter 2, page 17) 

Keysorters (Figure 3-5) are used to make a direct sort, or to sort cards 
into sequence. Two types are available: Keysorter-manual (single needle— 
adjustable extension); and Keysorter-manual (single needle—fixed ex¬ 
tension—supported at both ends)—recommended for sequence and single¬ 
needle selective sorting of cards 7x 8t£ inches and larger. 

The speed of sorting will depend somewhat upon the operator. The 
manufacturer states that 00,000 single hole sorts per hour are considered 
average, although 90,000 have frequently been reported. 

The alignment block (Figure 3-6) increases the speed and ease of 
sorting. The drop front fits flush against the front of the desk, which places 





















COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 33 




Figure 3-5. Keysorter, Types 5005 and 5006. 



Figure 3-6. Keysort alignment block. 


the block in the correct sorting position. The vertical guide along the right 
side of the block forms a right angle with the front edge of the desk. A 
rubber pad cemented to the bottom of the horizontal surface and the inside 
surface of the drop front prevents slippage during use. 

The key punch (Figure 3-7) is used for punching Keysort cards one 
at a time, one entire edge being punched at each trip of the operating lever. 




34 


PUNCHED CARDS 



Figure 3-7. Keysort keypunch, manual or electric. 



Figure 3-8. Keysort electric groover. 


This machine, available either in manually or electrically operated styles, 
is limited to outer edge single-row punching, and is primarily designed for 
numerical coding in the 7-4-2-1 fields. 

Grooving machines (Figures 3-8 and 3-9). As many as 100 cards, 
depending on the thickness of the stock, may be punched simultaneously 
in one position. This permits a considerable savings in time otherwise 
required for hand punching. 

Keysort Selector (Figure 3-10). This device, which handles cards up 
to 10)^ inches in length, makes a selective sort along an entire edge of the 




36 


PUNCHED CARDS 



Figure 3-11. Keysort Tabulating Punch. 


cards. About 200 cards can be handled at one time. A few simple selector 
codes suitable for use with this machine are described in Chapter 2. A 
reasonable speed is 60,000 multiple hole sorts per hour. When this selector 
is not in use, it may be folded and stored in a desk drawer. 

McBee Keysort storage cabinets are specially designed for use with 
Keysort installations. The drawers are light in weight, with handles front 
and rear, and the sides are cut down so that they can be used as utility 
trays when the system is in operation. 

Keysort tabulating punch (Figure 3-11) is a new 10-key adding and 
printing punch which automatically punches and tabulates quantities 
recorded in Keysort cards. It does a number of things: 

(1) Punches two quantities into the body of a Keysort card while simul¬ 
taneously printing such quantities on a detail tape and accumulating the 
amounts for totaling. 

(2) Automatically senses (reads) the quantities, punched as above, 
from the Keysort card and simultaneously prints such sensed quantities 
on a detail tape, thereby accumulating the amounts for totaling. 

(3) Automatically reproduce-punches quantities from a pre-punched 
Keysort card into a blank Keysort card, and again prints such reproduced 
quantities on a detailed tape, accumulating the amounts for totaling. 

(4) Totals the amounts accumulated in the machine and, as desired, 
summary punches such totals into a blank Keysort summary card. Prints 
such totals on the detail tape. 

(5) Can be used an an ordinary 10-key adding machine if desired. 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 37 



Figure 3-12. Keysort Data Punch. 


Keysort data punch (Figure 3-12) simultaneously prints data on a 
punched card and marginally notches the coded information into the Key- 
sort card. Metal plates of the Addressograph sort, embossed and notched 
with an employee’s name and code, for example, are inserted into the data 
punch and with one stroke of the lever the data are both printed and 
punched. 

The E-Z Sort System 

E-Z Soiit Systems, Ltd. 

45 Second St., San Francisco, Calif. 

In this system, all coding and punching is done in a half-inch strip 
around the perimeter of the card. Cards are available with a single row, 
double row, triple row or quadruple row of holes, or combinations thereof. 
The single-row hole card employs an oval-shaped hole. By staggering the 


38 


PUNCHED CARDS 


row, it is possible to place six holes per inch and eliminate any projection 
when adjacent holes are grooved. The multiple-row card effects its saving 
in space by using smaller holes and projections between adjacent coded 
holes are eliminated by the special pattern cut out by the punch or groover. 
The entire system may be handled manually, although electric punches 
and groovers are available. Four to six hours instruction, combined with a 
few days practice, will generally produce an efficient operator. Any E-Z 
Sort machine can be mastered within a couple of hours. All the equipment 
is portable and designed for simple operation. 

The system is intended for recording specific information on the card. 
Its expansion is unlimited, provided filing space is available. Present in¬ 
stallations vary from those using 10,000 cards per day to those using 10,000 
per year. The cards are durable; one installation has been in operation for 
over fifteen years. All types of E-Z Sort cards can be mended with a pat¬ 
ented glued strip which is claimed to make the repaired card stronger than 
before mending. 

E-Z Sort cards (Figure 3-13) are available in standard sizes ranging 
from 1 x 4 inches, in perforated strips containing four to sixteen cards, 
to sizes up to and including 8 x 10)4 inches, containing a total of 205 holes 
on single-row hole cards. Cards larger than 8 x 10)4 inches can be furnished 
by special processing. The number of cards that can be filed per inch 
varies from 80 to 150, depending upon the stock. A complete variety of 
card stocks can be obtained, including bond paper for sales tickets, etc., 
that are sorted once or twice and then filed or discarded, as well as special 
tag stocks and heavy durable rag content bristols for permanent records 



Figure 3-13. E-Z Sort card. 































COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 39 


and frequent sortings. E-Z Sort systems cards are also manufactured of 
special-purpose stocks such as Ozalid and other opalescent or sensitized 
materials. 

E-Z Sort’s patented hole structure, with depths up to four rows, is suited 
to direct-word coding with superimposed entries and is used extensively 
in research projects and technical literature files. Another feature is the 
ability to direct-sort letters or numerals up to four digits using a single 
sorter for each digit without incurring false drops. On three or four row 
holes, this particular coding arrangement uses fewer columns of holes than 
any other direct extraction arrangement. 

Other multi-row hole arrangements allow the coding of the largest num¬ 
ber of non-exclusive items in the least amount of perimeter space. 

A variety of standard stock cards are available for the small installation 
including five sizes of analysis cards with one or two rows of holes, five 
varieties of bibliographic index cards with two, three or four rows of holes, 
and several miscellaneous cards, such as forms control, mailing list control, 
and employment cards. 

Manual equipment (Figure 3-14) needed for the simplest type of 
operation consists of the hand groover, sorting needles and sorting tray 
which is used for holding cards while sorting. The hand groover permits 
the operator to code all four sides of the card at the rate of about 180 per 
hour. The hand groover is manufactured in several styles, including an 
intermediate type for cutting between the rows of holes on multiple row 
hole cards. The groover blade is replaceable when dull, and has a patented 
locator tip to assure proper register to the desired depth when coding. For 
example, the Model P-4 hand groover will groove one, two, three or four 
rows deep by inserting the locator tip into the hole being grooved. All of 
this equipment is offered for sale. 



Figure 3-14. E-Z Sort hand groover, sorting needles, and sorting tray. 



40 


PUNCHED CARDS 



Figure 3-15. E-Z Sort electric key-groover. 


The E-Z Sort electric key-groovers are available in two basic models: 
Model 1 (Figure 3-15) is a tabulating carriage type designed to groove 
7, 4, 2, 1,0 or other codings on the single-row hole card. With only seven 
keys all four edges of 500 to (500 cards per hour can be grooved. 

Model 2 (Figure 3-16) is designed to groove one entire edge of the card 
in a single operation. Keys can be set to repeat any part or all of the coding 
on succeeding cards without resetting the keys each time. This key-groover 
is manufactured in three styles—one row, two row or four row grooving 
depths. The two-row model grooves one or two rows deep and the four- 
row model grooves one, two, three or four rows deep or any combination, 
as may be required. 

Batch groovers are used to groove the same position in a large number 
of cards and are equipped with a locating pin to assure positive registry 
when grooving. These machines are manufactured in two models: electric 
(Figure 3-17) and foot-powered (Figure 3-18). The electric model is a desk 
type and the foot-powered model a pedestal type. They can be equipped 
with a special blade at the factory to groove one, two, or three rows deep. 
These machines are leased by E-Z Sort Systems, Ltd. 

The E-Z Sort Multi-sorter (Figure 3-19) is a selector unit. The needles 
are positioned according to the particular selective sort required. The unit 
may be set up in approximately one minute to separate cards coded in the 
required pattern along a single edge of the cards. It can handle 100 to 300 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 41 



Figure 3-16. E-Z Sort electric key-groover, for whole edge of card. 



Figure 3-17. E-Z Sort electric batch groover. 

cards at a time for a total of more than 36,(XX) cards per hour. Figure 3-19 
also illustrates the use of one of the catching trays. This device is available 
to handle any of the E-Z Sort Systems multiple-row hole cards up to 11 
inches long. 

E-Z Sort Card Counter. This machine is small enough to be placed 
on a desk and will accurately count approximately 1(X),000 3 x 5 inch cards 
per hour. It can be leased. 








batch groover 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 43 


The Unisort System 

Burroughs Corporation, The Todd Company Division 
Rochester 3, N. Y. 

Charles R. Hadley Plant, Los Angeles 12, Calif. 

UniSort is the registered trademark for The Todd Company line of 
edge-punched cards. UniSort cards are marketed through Todd-Hadley 
representatives in all parts of the United States, Canada, Alaska, and 
Hawaii. 

UniSort systems are manually operated and all phases of operation, 
including operation of the equipment, can be learned in a very short time. 
Maximum operating speed on the keyboard notching machine and maxi¬ 
mum sorting speed with the cards can be attained within a few days. 

All notching equipment, with the exception of hand notchers, is avail¬ 
able on a lease basis only. 

Standard UniSort cards (Figure 3-20) are available in sizes ranging 
from 3x5 inches to 6% x inches; special size cards can be made to 
meet the requirements of the system in which they are to be used. There 
is no standard UniSort card for bibliographic use. The card shown is a 
special one printed for a research institute. It illustrates triangular coding, 
so often used because it facilitates selective sorting. A deep hole notch on 
the left side and a shallow hole notch on the right, where the lines inter¬ 
sect, indicate the notching for the top letter or number in the square. Re¬ 
versing the deep and shallow notches indicates the notching for the bottom 



Figure 3-20. UniSort card. 








44 



Figure 3-21. U; 


t 






m ' ~ 



Figure 3-22. Uni 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 45 


The Keyboard notching machine (Figure 3-21) is used for notching 
up to three UniSort cards at a time, an entire edge being notched with 
one depression of the motor bur. Cards are fed into the front of the machine. 
This machine is available either in hand or electrically operated styles, 
for either four or five-hole-to-the-inch cards. 

The Foot-power notching machine (Figure 3-22) will notch the same 
hole in 200 cards at a time. Identical information (such as the date) may 
thus be notched into the edge of thousands of cards in a very short time. 

The desk model gang notcher (not illustrated) will notch the same 
hole in 50 cards at one time. This machine is used where the volume of 
cards is too small to warrant the use of a foot-power notching machine. 

UniSort alignment discs (Figure 3-23a) are used to hold a stack of 
cards in alignment while notching them on the foot-power notching ma¬ 
chine or the desk model gang notcher, thus attaining a more perfect notch 
for all cards in the stack. 

UniSort hand punches are available in four different styles: (1) ticket 
style (Figure 3-23b) for punching in the body of the card; (2) deep hole 
with gauge (Figure 3-23c); (3) shallow hole, with gauge and receptacle 
(Figure 3-23d); and (4) shallow hole, with gauge (Figure 3-23e). 



Figure 3-23. UniSort alignment discs, hand punches, and sorting needle. 



46 


PUNCHED CARDS 



Figure 3-24. UniSort card holder, shown on back of keyboard notching machine. 


The UniSort sorting needle (Figure 3-23f) is a lightweight, stainless 
steel needle set in a molded plastic handle. 

The UniSort card holder (Figure 3-24) slides into brackets on the back 
of the keyboard notching machine. It holds a stack of cards in such a 
position that the operator can read the top card and set the keys before 
removing the card to insert it for notching. 

The Sorting Pan (Figure 3-25) speeds the sorting operation. The raised 
guide along the right side of the pan stops the cards that have dropped out 
as those remaining on the needle are lifted over the guide. A rubber pad, 
cemented to the bottom of the pan, prevents it from slipping or from 
marring the desk. 

UniSort pull tubs, used as reservoirs for pre-punehed cards, are avail¬ 
able in two capacities, 9,500 or 15,000 cards. These are of steel construction 
mounted on rubber-tired casters. 

UniSort card cabinets are available for storing and filing UniSort 
cards. They are made of extra heavy steel, electrically welded, with a gray 
baked enamel finish. The cabinets contain either 8 or 16 drawers, each 
with a capacity of approximately 1,250 cards. 


COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 47 



Figure 3-25. UniSort sorting pan. 


The Findex System 

William K. Walthers, Inc. 

1245 N. Water St., Milwaukee 2, Wis. 

The Findex System is provided with cards having perforations equally 
spaced in the body and/or the edge. Space is provided at the top and on 
the reverse side of the cards for written records. Information is coded by 
cutting slots between adjacent vertical perforations. These slots are used 
in separating the cards by means of steel rods. 

Operation is entirely manual. The cards are typed, marked for slotting, 
slotted by hand, and filed vertically in any manner convenient to the user 
—for instance, alphabetically. They may be kept in an ordinary drawer 
or in the selector which is used when the cards are to be separated. The 
selector is a steel drawer, the front and rear plates of which contain per¬ 
forations matching those of the cards. Steel rods are inserted through the 
proper perforations, as determined by the method of coding, and extend 
the full length of the drawer. When the drawer is inverted and the cards 
stroked or shaken, cards in which there are slots corresponding to the rods 
drop a distance equal to the length of the slot. All cards are then locked in 
place by means of rods located at locking positions and the drawer is re¬ 
inverted. The selected cards may then be inspected or removed as desired. 
Each selector will accommodate 600 cards. It requires about three minutes 
to set the unit up for segregation; therefore approximately 12,000 cards 
may be sorted per hour. If the cards are not removed, no refiling is neces¬ 
sary. If they are removed, the refiling may be expedited by placing a colored 
slip in the file where each card has been removed. 


48 


PUNCHED CARDS 


No particular skill or training is required. Anyone who is capable of 
typing the cards can handle the sorting and other operations after reading 
the operator’s manual supplied by the manufacturer. All equipment is 
sold outright but may also be obtained on a rental basis, subject to sale 
within a limited period. No special maintenance problems are involved. 
The expansion of this system is limited only by the amount of filing space 
available. However, expansion relative to the amount of information to 



Figure 3-26. Findex card. 




COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 49 


be coded or recorded on the cards is limited by the size of the card. Cards 
(Figure 3-26) are available in two sizes—6 x 8 and 8x8 inches. They are 
made of 0.012-inch ledger stock with a high linen content and are durable 
enough to outlast ordinary usage. Wrong coding may be corrected by 
pasting a small linen “patch” over the slotted portion between the two 
holes. The intended use of the card will determine how much of its area 
is to be used for punching and how much is to be used for written records. 
No provisions have been made at present for microfilm insertions or photo¬ 
sensitization. 

Selectors. The selector drawers, which may also act as permanent filing 
space, house 600 cards each. They may be handled singly on a revolving 
“cradle” or in cabinets of two, four, six, or eight selectors each. These 
cabinets are provided with special slides which permit inversion of the 
drawers during the selecting process without removing them from the 
cabinet. A cabinet holding eight such drawers (4,800 cards) occupies three 
to four square feet of floor space, depending on the size of the card. 

The only other implements required are the steel sorting rods and the 
slot punch which resembles a desk-type paper punch. All of the equipment 
is portable. The system appears relatively inexpensive and well adapted 
to the requirement for a large number of cards and a comparatively small 
amount of information on each card. It is primarily designed for records 
which require group analysis, correlation, and cross indexing, when it is 
desirable to have all information on one card. It does not lend itself to 
tabulation of accounting facts or to serial sorting. 

The Flexisort System 

Superior Business Machines, Inc. 

285 Madison Ave., New York, N. Y. 

The Flexisort system operates upon the same basic principles as the 
marginal hole punched card systems described previously but it is unique 
in its use of a punch which automatically cuts all necessary holes into un¬ 
perforated cards. Such a machine eliminates the need for specially prepared 
cards. Any kind, size, or weight card stock can be used. It also permits 
conversion of existing index card files into marginal hole punched-card 
files. 

Flexisort cards (Figure 3-27) may be prepared either singly, as one of 
the parts of a set of precollated snap-out forms, or they may be prepared 
in continuous form. When prepared in continuous form, it is possible to 
type an invoice or listing as the original copy and simultaneously produce, 
as a carbon copy, a separate card for each entry on the list. 

Flexisort punch (Figure 3-28). This machine is electrically operated 
and resembles a desk-type adding machine with an eight-bank keyboard. 



50 


PUNCHED CARDS 




Figure 3-28. Flexisort electric punch. 


Each bank contains a correction key, numerical keys numbered 1 through 
1), a repeat key, and four supplementary keys which permit the direct 
coding of miscellaneous statistical information. Every second column of 
keys carries alphabetical characters as well as numerals. The cards are 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 51 


perforated and punched by depressing the keys corresponding to the posi¬ 
tions to be coded and then pressing down the motor bar. The single stroke 
perforates the holes and punches the notches along one edge of the card. 
All eight columns may be set prior to the simultaneous cutting operation. 
When a numerical key is pushed, four dies are arranged within the machine 
and when activated they punch a pattern into the margin of the card 
corresponding to the 7-4-2-1 code previously described. Each number 
from 1 to 9, inclusive, can be punched with a single key depression, but 
for alphabetical punching two keys are depressed for each letter. For 
example, to code “B,” the key marked with the letters “ABC” and the 
key “2” (signifying the second letter of that group) in the column imme¬ 
diately to the right are both depressed. Each of the keys provided for direct 
coding controls only one die and may not be used simultaneously with the 
alphabetical or numerical keys in the same column. Erroneously set keys 
may be released by depressing the correction key. The repeat keys lock 
the dies into position and make it possible to cut common information 
into successive cards without resetting the board. As many as three cards 
can be gang punched simultaneously. Provision is made for suppressing 
punching in areas where it is not desired. 

The Needlesort System 

Arizona Tool & Die Company 
31 E. Rillito St., Tucson, Ariz. 

Needlesort cards (Figure 3-29) are manufactured in two sizes, 3)^ x G 
inches and 5x8 inches. They are only punched along three edges—the 


Hi 



rim 

llfl 

ntnimin 

i 

gj 

# N 

SHOP SERIAL: 

MISSILE/COUIPMI 

.NT TYPE 

FAILED IN OR 

n 

DUE TO: DSTO 

□ HAN 

INSPECTION 

RAGE 

DUNG 

IMPTED LAUNCH 

HT 

A 

Si 

# * 

OATE 

MISSILE/EQUIPMENT SERIAL: 

U ATTI 

n TEST/CHECK □ FLIC 

# N 

• « 

WHAT WERE THE 

SYMPTOMS OF FAILURE? 

ESTIMATED ACCUMULATED OPERATING TIME 

OF COMPONENT AT FAILURE 




PLATES ON' 

MINS 

k 

FILAMENTS “ON" 

HRS. 

ENERGIZEO 

HRS. 


• ts 

ITEM: 

NAME. MOO EL. OR TYPE: 


MEASURED OUTPUT OR CONDITION | 

ts 

BEFORE REPAIR 

AFTER REPAIR 

B 

# ' l 

• « Z 

COMPONENT OR 
UNIT 





i 

• * E 

SUB-ASSEMBLY 

OR CIRCUIT 





B 

u 

# n * 

INDIVIDUAL PART 
OR ELEMENT 






WHAT WAS WROA 

G WITH THE PART? 


B 

• W 

• o 5 

#0 6 

• « s - 

• < < 

| WHAT DIO YOU DO TO REPAIR THE COMPONENT? 

Je 

PROBABLE CAUSE OF FAILURE (REMARKS OR RECOMMENDATIONS! 



MAN HOURS REOUIRED 1 □ SfC Otmcr Sioc For CHECKEO AND APPROVED BY 

OATE: 

m 

I DEFECT CARD: 

IB 


Figure 3-29. Needlesort card. 























52 


PUNCHED CARDS 


two sides and the top. The 5x8 inch cards are available in three colors, 
buff, pink or yellow, and can be imprinted or mimeographed by the com¬ 
pany. A notching punch and sorting needle are also offered for sale. 

The Zatocoding System 

Zator Company 

lJtOYi Ml. Auburn St., Cambridge 88, Mass. 

The Zatocoding System consists of the Zator Selector, edge-notched 
Zatocards, the techniques of using random descriptor code patterns notched 
in superimposition along the edge of the cards, and the use of descriptors 
by which documents are characterized. Deriving descriptors is an empiri¬ 
cal process. Chapter 15 describes this process for a given installation. 

The Zator 800 Selector (Figure 3-30) holds an easy handful of Zato¬ 
cards. The box is vibrated by a small motor. Rods or needles running from 
front to back of the box are inserted in a desired selective pattern. Cards 
notched in the positions corresponding to the selector rods fall down from 
the rest of the pack and can be easily separated. With the 800 Selector, 
sorting speeds of about 800 cards per minute can be attained. 

Zatocards come in two styles, one with notches along a single edge and 
the other with notches along two edges (Figure 3-31). The latter type has 
72 notching sites as compared to 40 on the single-edge style. 

For the Zatocoding system, a list of random-like patterns has been 
prepared. Any patterns that are random-like, in the sense that the individ¬ 
ual code marks are well scattered and fall with approximately equal in¬ 
cidence on all the coding sites, can be used for coding. 



Figure 3-30. Zator 800 Selector. 




COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 53 



Figure 3-31. Zatocard. 


Technical consultation and guidance are provided by the Zator Company 
during the installation of a commercial Zatocoding system. The system is 
provided on a rental basis, including a license for making the randomly 
coded cards. 


Foreign Manufacturers 

Edge-notched punched cards are manufactured by a large number of 
firms outside the United States. The list that follows is by no means ex¬ 
haustive. 

Buro-Organisation G. m. b. H. Esselte System 

Brandenburgische Str. 27 Tandlarkarhogskolan 

Berlin W 15, Germany Malmo, Sweden 


Copeland-Chatterson Company, 
Ltd. 

Exchange House 
Old Change 

London E. C. 4, England 

Edler & Krische 
Kestner Str. 42 
Hannover, Germany 


National Luchtvaartlaboratorium 
Amsterdam 
Sloterweg 145 

Amsterdam-W., Netherlands 
Rapidtri 

78, Rue de Wattignies 
Paris 12®, France 


Eichhoff-Werke G. m. b. H. 
Dieffenbach Str. 2 
Schlitz/Hessen, Germany 


G. Schmid Verlag 
Herderstrasse 2 
Lubeck, Germany 




54 


PUNCHED CARDS 


TABULATING TYPE PUNCHED CARDS 
The IBM System 

International Business Machines Corporation 

590 Madison Ave., New York 22, N. Y. 

IBM accounting machines are widely used not only in commercial account¬ 
ing and statistics but also for processing many kinds of scientific data. 
In laboratories throughout the country, these machines are used to create, 
maintain, and search punched card reference files and to perform mathe¬ 
matical computations necessary in solving many problems in astronomy, 
ballistics, chemistry, engineering, meteorology, and physics. 

Once the initial data have been recorded as holes in the cards, the ma¬ 
chines can sense these holes electrically and automatically perform a wide 
variety of operations such as rearranging the cards into any required se¬ 
quence, transferring data from one card to another, printing the informa¬ 
tion on the cards or on a sheet of paper, consulting tables of data, and 
performing the arithmetical operations of addition, subtraction, multipli¬ 
cations, and division. Electrical impulses, transmitted through the holes 
in the cards, are used to read the recorded data and control the operation 
of the machines. 

Branch offices, service bureaus, and representatives are located in prin¬ 
cipal cities. Equipment can be bought or supplied on a monthly rental 
basis which includes the use and maintenance of the machine. For those 
who do not have the necessary installations, the Service Bureau Corpora¬ 
tion, a wholly owned subsidiary corporation, offers sendees for the prep¬ 
aration of reports on a time or complete job basis. 

The company maintains a training program for the representatives of 
its customers. Classes are held at IBM offices and educational centers for 
instructing the customer personnel in key punching, operation of the vari¬ 
ous machines, control panel wiring, and other related subjects. Selection 
of the personnel to receive instruction is determined by the customer. 
Classes also are held for superv isors and managers of machine accounting 
departments. At Endicott and Poughkeepsie, New York, IBM conducts 
classes for customer executive personnel. 

Cards (Figure 3-32) are supplied in a variety of sizes but the most widely 
used one is Z x /i x 7 3 g inches. A stack of approximately 150 cards measure 
one inch in thickness. The 3 ) 4 x 7 3 g inch card contains 80 vertical columns 
divided into 12 punching positions. Combinations of these 12 punching 
positions are used to punch alphabetical and numerical information in 
the card. 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 55 


ABCDEFGH1JKIMNOPQRSTUVWXYZ l*»«9G78»0 

lllllllll 

■IIIIIIII 


AlPHAGEYICA l, P RINTING PUNCH 

i mi in ill ii 


n 


in i i ii 


imm • • nan tta 
1111111|11111 

2 t22t222|2222 

S2)2S>Sf)|2)) 

4 4 4 4 4 4 4 4 4 414 4 

SSSSIIIIfSS|I 

lllllllllllll 

7777777777777 

• aiiiiiiiiiu 
i •nnn i mi 

i i i* • i i it anooi* 


ii in 11111111 ii it in |iiii ii in ii |ii i|n m|ii ii in ii ii n lllllllll 
222 | 2222222 | 222222222 | 2222222222222 > 2|222222222222222222222222222 
ISSS|SISSS!S|ltSStlSSS|StS)SSStSf|IISSI|»|S|tt!IS|S2ISSII|SIIIIII 
44444|4444444|444444444|4444444444444444444444444444444|444444444 

iiisis|s suns |s»iiissii|isis9sssss>ss|issiiiisi|si|ss»»|i»s»i2ii 
i n lll•|ll••l•l■n ill ii ii|ni ii mi ii iiiiniiii mi iiiiiiiii mil 

|77777777|7777777|777777 77 7|7777 777|7777 777777|77H77|7|77 77777777 

..... 

III!”II” li!” I” »”!!!!” 


Figure 3-32. IBM card. 



Figure 3-33. IBM card punch, Type 24. 


Key Punches are used for punching the holes in the cards. A number of 
different machines can perform this function. The card punch (Figure 3-33) 
is used for recording both alphabetical and numerical information, and is 
equipped with a combination keyboard designed for high-speed operation. 
Cards are fed into the machine automatically and move forward column 
by column under the control of a program card which governs duplicating, 
skipping, and the kind of information (either alphabetical or numerical) 
to be punched into the cards from the combination keyboard. A duplicating 






















50 


PUNCHED CARDS 




% m?m\ v 4 tt— 

REG 

(ESDI 

|I„aJ 



i—-iMn.i 

■BlkM 


JO» TIME CARD 


□ * 


□ * 
□ ’ 


DEC 31 10 .5 


DEC 31 8 .0 


1 


STOP 


START 



ACCOUNT OR JOB NO. 




Jc7^c7^c7^7^c7^7- 

Pr 


4*H P7 Tni] 


PART NO. q OPCR NO. MACN NO. PIECES FINISHED HOURS 

JPdP&c&dP#* 
cPcP^d^cl-cPcPcPcl^crl^rP 
dPcT? dPdP dP&dPdP 




i 

5 5 c5 : ^c:5 ? |5jc5^c| 3 c5 3 c5j=5 3 c5jc5 3 c:5^c5 :5 c5 ;:) ȴ e 5 juSl 


&c&c1P^ctPtePciP^c1Pc&ci>~ c6=(c6^ 
TPcPcP €PcP cUcttUcl^cUcU e7=je7=| 
diP ciPciPdiPciPdiPciP c0:S 

■ 9 >9 i 9'V-9^ .: 9 I 9 ; 9 -'^, 

jffiE 


miiiMiunmyrniiniirnini 


R 


:«=fc«= 

i=U-*i 

il. 



Figure 3-34. IBM “Mark-sensed” card. 


feature permits automatic punching of common information from one 
card into the next. A printing card punch is available that performs all 
the functions of the card punch and which also prints along the top edge 
of the card the characters that are coded in its columns. 

Cards can also be punched automatically, without the use of key punch¬ 
ing, through a medium known as “mark-sensing”. This embodies the use 
of a graphite pencil to make short marks directly on the face of the card. 
The cards marked in this manner (Figure 3-34) are fed into another ma¬ 
chine which electrically senses the graphite marks and punches correspond¬ 
ing holes into the desired position on the same card. 

Still another way to enter original data into cards is by means of the 
Typewriter Card Punch, which will simultaneously type a document 
and punch cards with selected data from the typewritten record. IBM 
also manufacturers a typewriter tape punch which records selected data 
into a paper tape. This tape can be transmitted by mail or teletype to a 
remote location where a tape-to-card machine will convert paper tape to 
punched cards. 

Verifiers detect transcription errors. Once the card is punched, it be¬ 
comes the basic record from which all transcription is subsequently done 
by machine. However, since this does not relieve the possibility of errors 
in the original punching, the verifier has been provided to check punching 
accuracy. As in key punching, the verifier keys are depressed as if the same 
information were being recorded once more. If the punched positions in 
the card do not correspond to the keys depressed in the verifier, an error 
signal is made. It can then be determined whether the error is in the punch¬ 
ing or in the verification. Cards are notched on the right hand edge as 
visible proof that they have been verified and are punched correctly; in 
























COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 57 



Figure 3-35. IBM sorter, Type 083. 


the event of an error they are notched directly over the column in which 
the error appears. 

A device known as the Self-checking Number Device used with the Key 
Punch may, in some cases, obviate the use of a verifier. With this device, 
the Key Punch will assign one extra digit to any code or identifying num¬ 
ber. Thereafter, when that number is used, errors either in the original 
handwritten record or in the punching of the card are automatically re¬ 
vealed. 

Sorters (Figure 3-35) arrange cards in any desired order according to 
the data punched into them. Also they separate the cards into groups 
having certain specific information. This sorting process takes place one 
column at a time, so that by successive sortings the cards may be arranged 
in any desired order. The 13 pockets into which cards can be distributed 
correspond to the 12 vertical punching positions in a card plus one pocket 
for cards having no holes in the column being sorted. If any given data 
are being selected, this last pocket receives those cards not having the de¬ 
sired information. The cards deposited into any one pocket remain in the 
same sequence in which they were fed into the machine. These units auto¬ 
matically stop when a pocket is filled. The one illustrated handles 39,000 
sorts per hour. 


PUNCHED CARDS 


58 



o 


1 ! ^ 

Figure 3-36. IBM collator, Type 077. 

Collators (Figure 3-36). The principal function of the collator is to feed 
and to compare two sets of punched cards simultaneously, in order to 
match them or to merge them. While doing this the collator can separate 
the cards which match from those which do not, thereby making it possible 
to pull as well as to file cards automatically. There are two feeds, each 
of which operates at the rate of 14,400 cards per hour, and there are four 
pockets into which the cards can be separated, for example, into two groups 
of matched cards and two groups of unmatched cards. As a filing machine, 
the collator simultaneously feeds and compares two groups of cards. These 
two groups are merged into a single group in numerical or alphabetical 
sequence, or, if preferred, the cards may be matched into two identical 
groups. While either of these operations is being performed, the collator 
will remove from either group those cards that do not match, those which 
are out of sequence, or those which match cards in the other group or 
other selected cards. 




COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 59 



Figure 3-37. IBM alphabetical accounting machine, Type 403. 


Printing Units. The IBM accounting machine is used to obtain printed 
reports of data punched in the cards. This is a machine which selects and 
reads data from cards, adds and subtracts, and prints on a sheet of paper 
data from individual cards or from accumulated totals. The machine will 
add or subtract as many as 112 digits at a time and print as many as 120 
characters in a single line. There are several types of accounting machines, 
all basically the same in operation. Some of these print only numerical 
data, while others print both numerical and alphabetical data. Several 
models will print three lines from a single punched card. 

Most widely used, however, is the machine (Figure 3-37) which prints 
one line of numerical or alphabetical information from each card. This 
unit will print 88 positions of numerical data, or 43 positions of alpha¬ 
betical data on the left, and 45 positions of numerical data on the right. 
Every character on a line is printed simultaneously. The unit may be used 
to print selected details from every card or from specific cards. It also 
may be used to accumulate selected data and print the various classifica¬ 
tions of totals. While several speeds are available, the fastest will perform 
detail printing at the rate of 9,000 lines per hour and will accumulate 
totals without detail printing at the same speed. The machine is equipped 
with major, intermediate, and minor classification controls which provide 
for the printing of group totals when any of these classifications change. 
Important to the high speed of this operation is the tape-controlled car- 




60 


PUNCHED CARDS 


riage, which automatically feeds continuous forms to the correct printing 
positions. 

The accounting machine, like most others in the IBM system, is pro¬ 
vided with a control panel on which pluggable connections can be made by 
the operator. In this way, for example, the number read from a given card 
column can be routed to any adding or printing unit. Other connections 
determine the operations to be performed as each card passes through the 
machine. The control panel can be easily removed for wiring changes or 
for replacement by another panel that has already been wired for a differ¬ 
ent type of operation. 

Summary Punches. To record totals accumulated in the printing unit 
into other cards for use in subsequent operations, a punching unit can be 
attached to the accounting machine by means of a cable. There are a num¬ 
ber of summary punches, several of which are also used as key punches; 
one goes even further and performs five different machine functions, thus 
reducing the number of machines necessary to handle relatively light work 
loads. Another, the Accumulating Reproducer, is used for accumulating 
totals and punching summary cards independently of a printing unit. 
This machine checks all of the accumulated totals as well as their punching. 

Reproducers automatically transcribe punching from one card to an¬ 
other, thereby limiting clerical transcription to the original operation of 
key punching from source records. The most flexible machine which per¬ 
forms this as well as other functions is the Accumulating Reproducer. 
Reproducing in new cards need not be done in positions corresponding to 
those of the old cards. The positions can be selected and controlled by a 
removable and flexible control panel quickly inserted into the machine in 
much the same manner as in the Accounting Machine. This machine also 
will select and reproduce only those cards desired, without disturbing the 
arrangement of a file. 

The Accumulating Reproducer will reproduce punching from one 
card into all cards following it until a new “master” card passes through 
the machine, instructing it to reproduce information appearing in that 
card. This is known as “gang” punching. These two forms of reproducing 
can be done simultaneously; that is, while information is being reproduced 
from one to another group of cards, additional data may be punched by 
means of one or more master cards. Both of these operations, working 
independently or simultaneously, take place at a speed of 6,000 cards per 
hour. 

Other functions of the Reproducer include: printing in large size type, 
from data punched into the card, as many as eight figures on the edge of 
the card; comparing the reproduced and reproducing groups of cards to 
verify the accuracy of the reproduction; punching cards which have been 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES G1 



Fig. 3-38. IBM calculating punch, Type 602-A. 


marked-sensed by a graphite pencil; and performing summary punching, 
as described previously. Summary punching and reproducing can be done 
simultaneously. 

Calculating Punches. (Figure 3-38). To perform the routine as well as 
the complex calculations encountered in commercial and scientific work, 
IBM has five punched-card calculators, all varying in capacities and speeds. 

One of these machines is the Calculating Punch. As cards pass through, 
this unit reads the factors, adds, subtracts, multiplies, divides, and punches 
the results. A multiplicand punched in the card can be multiplied by a 
factor in the same card or by a group multiplier punched in a single master 
card. There are several methods of controlling the machine to check the 
final results. 

Several independent problems can be performed on this machine and 
several results punched in the card. The multiplier may contain as many 
as eight digits, the product as many as 30. Multiplying speed varies with 
the size of the factors and may be as high as 3,000 extensions an hour. 
The machine has a capacity for an 8 position divisor, a 15 position divi¬ 
dend, and an 8 position quotient. Dividing speed, as in multiplying, varies 
with the size of the factors to as high as 1,000 divisions an hour. 

Various series of basic operations can be performed in any sequence as 





02 


PUNCHED CARDS 



Figure 3-39. IBM electronic calculating punch, Type 604. 


a card passes through the machine. Different factors may he added and 
the sum used as a multiplier, multiplicand, dividend, or divisor. A product 
or quotient can he further multiplied or divided hy additional factors, and 
amounts may he added to or subtracted from the calculated results, the 
result being punched in the card in the same operation. Speed depends 
upon the complexity of the problem and the size of the factors. 

The Electronic Calculator, Type 604 (Figure 3-39) is used when 
greater speed is essential. Calculations performed by this machine are 
made at a rate of 0000 cards per hour regardless of the operations involved. 
This machine functions in much the same way as the calculating punch 
described above. A point of difference is the fact that the calculating and 
punching operations are performed by different units. When even greater 
capacity and speed are important, the Type 007 Electronic Calculator may 
be used. 

The Card-programmed Electronic Calculator. For problems re¬ 
quiring long series of arithmetical operations to obtain a single solution, 
the Electronic Card-programmed Calculator is widely used, since it per¬ 
forms these operations automatically. This ability, in addition to its stor¬ 
age capacity, high-speed computing, and high-speed printing, makes this 
calculator especially useful for the more complex problems encountered in 
engineering, statistical, and scientific work as well as in commercial applica¬ 
tions. 

A later development, the Magnetic Drum Data Processing Machine, 






COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 03 



Figure 3-40. IBM electronic statistical machine, Type 101. 


employs more advanced storage techniques that make it possible to process 
problems requiring internal storage of as many as 20,000 decimal digits. 

The Electronic Statistical Machine (Figure 3-40) combines in a 
single unit the functions of sorting, counting, accumulating, balancing, 
editing, and printing of information. Unit counts may be distributed into 
as many as ()0 different classifications while the basic data are sorted at 
the rate of 450 cards per minute in any desired sorting pattern to provide 
for further cross classifications. During the same run, information in the 
cards may be automatically checked on the basis of pre-established criteria 
for consistency. Files of cards may be searched automatically for specific 
numbers or ranges of numbers. By eliminating intermediate card handling 
and processing on other machines, the statistical machine provides a means 
for obtaining comprehensive statistical analyses in a relatively short time. 

One of the most important developments in the field of high-speed data 
processing is IBM’s Type 709 Data Processing Machine. Made up of vary¬ 
ing numbers of units, depending on the work to be performed, the out¬ 
standing characteristics of the “709” are its very large storage capacities, 
and its very fast reading, writing, and computing speeds. The principal 
contributions to its speeds and to its flexibility in processing many types 
of problems are made by its three advanced types of storage—magnetic 
tapes, magnetic drums, and magnetic cores. As an example of the over-all 
speed of the “709”, it is capable of performing up to 42,000 mathematical 
operations a second. 

Input of data is by means of punched cards or by magnetic tapes. Out- 


64 


PUNCHED CARDS 


put is in the form of reports printed at a speed of 500 lines a minute, punched 
cards, or magnetic tapes. 

The Remington Rand System 

Remington Rand, Inc. 

315 4th Ave., New York 10, N. Y. 

The Remington Rand punched-card accounting method is based on 
punching holes in 90-column tabulating cards to code information perti¬ 
nent to each business transaction or subject of interest. The holes in the 
card actuate machines that sort, add, subtract, multiply, punch, collate, 
interfile, and prepare in printed form records and reports in accordance 
with a company’s over-all record keeping requirements. 

In making an installation, the supplier usually selects persons employed 
by the customer and trains them in the operation of the equipment. The 
training time for each machine varies from a few hours to several days. 
The machines which have a keyboard like a typewriter demand the devel¬ 
opment of some skill, and require from two to three weeks for the operator 
to acquire a moderate amount of speed. However, several hours of in¬ 
struction usually enable one to understand the operation of most machines. 

Trained mechanics are located in major cities to service the equipment 
and to keep it in good operating condition. The machines are either sold 
outright or leased on an annual basis. When these are leased the rental 
includes service, but for machines sold outright there is available a me¬ 
chanical service agreement under which, for an annual fee, Remington 
Rand agrees to keep them in operating condition. The purchaser may 
assign one of his own representatives to be trained in the maintenance of 
the machines at the supplier’s mechanical training school. Parts are sup¬ 
plied at a nominal cost. 

Tabulating cards (Figure 3-41) measure x 1% inches with a thick¬ 
ness no more than 0.007 inch or less than 0.00625 inch. These cards are 
manufactured to exacting specifications from a high quality paper stock 
and are printed to meet the needs of the specific application. 

On the card illustrated there are 540 punching positions divided evenly 
among 90 vertical card columns, 45 on the upper half and the same number 
on the lower half. The punching code used for the six punching positions 
provides for recording 37 characters, 0 through 9, and A through Z, and 
one special character. Only one of the 37 characters may be coded per 
column, except under special conditions where each of the six positions in 
a column may be wired to a specific character. 

Automatic punches (Figure 3-42) operate by simultaneous perforation 
on the punch die principle. This feature, exclusive with Remington Rand, 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES fi.5 



Figure 3-42. Remington Rand automatic punch. 
























06 


PUNCHED CARDS 


provides for setting of the dies in the punch by depressing the keys and 
for perforating the entire card with a single depression of a “trip” key. 
This enables the operator to correct an error before the card is punched. 
Also information common to each card may be recorded into a series of 
cards without resetting the keyboard for each card. Two basic key punches 
are available to cover the general range of manual punching requirements. 
One of these is for numerical punching only and the other for both numeri¬ 
cal and alphabetical punching. Devices and attachments are available to 
modify these basic machines for particular applications. In addition, both 
machines may be used for “verify” punching as well as for the original 
punching entry. The production capacity of these machines varies with 
the design of the tabulating card and the type of information to be punched 
in each card. It is claimed that the average operator can punch 1500 90- 
column cards in an 8-hour day, but this may vary from a few hundred to 
as high as 5400 an hour depending upon the amount of variable information 
punched into each card. 

Automatic Verifying Machine. To verify the correctness of punching 
in a pack of cards, a second operator repeats the original punching opera¬ 
tion with a control on the automatic punch set for verify. This operation 
elongates all correctly punched holes and leaves round holes where there 
are errors. The cards are then taken to the Automatic Verifying Machine 
which mechanically senses the verify-punched cards. This machine places 
a card of contrasting color having uncut corners on top of each card con¬ 
taining a round hole. It also punches a small hole in the right-hand margin 
of each card passing through the machine to show that it has been verified. 
This machine has a constant production speed of 200 verify-punched 
tabulating cards per minute. For standard card-per-card interleaving the 
speed is 400 cards per minute. 

The Sorting Machine embodying mechanical principles of reading the 
holes punched in the cards is known as the Standard Sorter. It operates 
automatically at a speed of 420 cards per minute. It requires only the in¬ 
sertion of the cards and their removal in sequence. The machine stops 
automatically when the last card is fed out of the feeding magazine, when 
a card becomes wrinkled or damaged, and when the receiving magazine or 
magazines are filled to capacity. One operator can handle up to four sort¬ 
ing machines depending upon the complexity of the operation. Additional 
devices are available for the Standard Sorter which permit group sorting 
(searching), and an operation which is known as “pairing,” and counting. 

The Electronic Sorter, (Figure 3-43) a new addition to the Remington 
Rand punched-card line employs the principle of “black light” reading of 
the card; it operates at the high speed of 800 cards per minute. Different 
from the standard sorter, this machine is equipped with 13 receiving maga- 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 67 



Figure 3-43. Remington Rand electronic sorter. 


zines and one reject pocket which permits alphabetical sorting with one 
and one-half passes of the cards through the machine. In other words, on 
the first pass through the machine, the A-M letters are sorted in exactly 
that sequence and after clearing the receiving pocket, the remainder of 
the deck containing the N-Z letters are put through the machine and these 
letters are sorted in exactly that sequence. Any numerical punching which 
may occur during either of these passes will automatically fall in the reject 
pocket. The Electronic Sorter embodies all of the safety devices in the 
form of automatic stoppings that are included on the Standard Sorter 
described above. 

Printing Tabulators (Figure 3-44) transpose the punched information 
from the card into a printed report. They are of two types—numerical 
and a combination numerical and alphabetical. All alphabetical tabulators 
have type bars which print any numerical figure from 0 through 9 and 
letters from A through Z and one special character. The large capacity 
alphabetical tabulators have up to 100 sectors and will print 100 characters 
simultaneously. Counters for the addition or subtraction of numerical 
information can be installed on all alphabetical tabulators. Two counters, 
one to give totals and the other grand totals, can be attached to each type 
bar. Each of 80 type bars can be equipped with two of the direct subtrac- 



68 


PUNCHED CARDS 



Figure 3-44. Remington Rand alphabetical tabulator. 

tion type counters, thus providing 160 counters for adding and subtracting. 
Output is at the rate of 6,000 machine cycles per hour. A machine cycle 
consists of printing and accumulating one tabulating card or printing one 
total from a group of tabulating cards. 

The Punched-card Interpreter prints across the face of a tabulating 
card whatever alphabetical and numerical information is punched into the 
card. The location of each printed character is normally directly above the 
zero position in the column in which it is punched, but, if desired, it may 
be placed on any one of thirteen printing positions on the upper half of 
the card, at a speed of 90 cards per minute. 

The Posting Interpreter makes possible taking information which is 
punched in one card and printing it on a following card. Two important 
applications are the preparation of tabulating card checks and the posting 
of employees’ earning records on tabulating card ledgers from information 
punched in weekly net earnings summary cards. 




COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 69 


The Posting Machine is the automatic line finding posting machine 
which permits the use of tabulating cards as ledgers for historical, chrono¬ 
logical, or continuous listings of transactions that occur within an account. 
Prior to the posting operation, the punched cards created either by key 
punching or summary punching are collated with the ledger cards and the 
posting machine reads the accounting information punched in the leading 
card, automatically selects the next open line for posting on the ledger 
card and prints the information on the ledger. This posting machine 
operates at a speed of 90 cycles per minute which gives a net productive 
posting speed of 45 postings per minute. It is equipped with a dual-card 
receiving magazine so that the punched detail cards are automatically 
segregated from the printed ledger cards without going through a subse¬ 
quent machine operation to separate the two sets of cards. 

The Multi-control Reproducing Punch reproduces information from 
one tabulating card into an unpunched card. Its function is to compare 
two cards with respect to their coding and then to reproduce the variable 
data from the primary (master) card into one or more secondary (detail) 
cards, depending upon the existence of a matching or non-matching condi¬ 
tion. At the same time cards that match can be segregated from cards that 
do not match. The output of this machine is from 100 to 200 cards per 
minute depending upon the operation being performed. 

The Interfiling Reproducing Punch, an improvement over the 
Multi-control Reproducing Punch, is used for comparing two decks of 
cards and for punching, segregating, or interfiling, depending upon the 
results of the comparison. Its speed is identical to that of the Multi-con¬ 
trol Reproducing Punch. 

The Collating Reproducer, (Figure 3-45) a modification of the Inter¬ 
filing Reproducing Punch, automatically compares, punches, interfiles, 
segregates and sequences 90-column tabulating cards. Thus, it can be used 
to verify, group-extend, code, decode, in-file, and out-file. Any errors 
found by this machine when checking the sequence in a single card file 
are marked by signal cards and the erroneous cards can either remain in 
the file or be removed. Two separate card files arranged in numerical 
sequence can be interfiled. This machine also controls the feeding of cards 
from one or more feeding magazines in numerical sequence. Segregation 
and selective punching can be accomplished simultaneously. These opera¬ 
tions are performed at rates of 6,000 to 12,000 cards per hour. 

The High-Speed Electric Collator (Figure 3-46), operating at a speed 
of 250 machine-cycles per minute, embodies the electric principle of sensing 
cards. The collation is controlled by a flexible wiring panel which can be 
readily changed by the operator; or extra pre-wired panels are available 
which can be readily installed to change the setup. A feature of the Elec- 



70 


PUNCHED CARDS 



Figure 3-45. Remington R:uul collating reproducer. 

trie Collator is its ability to check the sequence of both sets of cards being 
fed from the two feeding magazines without loss of its ability to merge or 
segregate in the four receiving pockets. Although the machine is basically 
a numerical unit, it nevertheless will compare alphabetical information 
punched in the two files and will segregate cards which do not match. 

The Calculating Punch performs the arithmetical operation of addi¬ 
tion, subtraction, multiplication and division from values sensed from 
punched cards. The results of these operations are then punched into the 
same card from which the values were obtained or into any card which 
follows. 

In arriving at the final result or results for an application, the machine 
follows a pre-planned course of operation for the processing of one card 
through the machine. This is called a program. The individual elements 
of the program for any given application are program steps. Provision is 
made for 12 program steps. In any one application, part or all of the steps 
may be followed as required. It is also possible to expand the program by 
repeating steps. Within one program, two or more subroutines (sub¬ 
program) involving one or more steps may be followed. The Calculating 
Punch will also add or subtract two or more values sensed from the same 
card to arrive at one or more results. The result, or results, thus obtained 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 71 



Figure 3-46. Remington Rand electric collator. 


may be used for additional calculations for the same card or the following 
card. This is called “cross-footing.” 

A feature of the calculating punch is that values arrived at early in the 
program for one card may be punched into that same card while the ma¬ 
chine proceeds with further calculations for the same card. On the other 
hand, the machine will also proceed with the calculations for one card 
while simultaneously punching the result into the card immediately pre¬ 
ceding it. 

The entire machine performance for individual applications is obtained 
through, and is controlled by, the panel. By means of this panel all phases 
of machine operation may be varied from application to application to 
meet the individual requirements. The operator may pre-wire this panel 
as required for individual application, or he may readily install a pre-wired 
panel for the new or subsequent application. 

The Punched Card Electric Computer (Figure 3-47) performs the 
arithmetical functions of addition, subtraction, multiplication and division 
of values sensed from punched cards, values manually set into the machine, 
or values computed during a sequence of programming a problem. Results 




72 


PUNCHED CARDS 



Figure 3-47. Remington Rand punched card electronic computer. 


of these functions are then punched into the same card from which the 
values were sensed, or into any desired card which follows, or they are 
placed in storage for subsequent use in the computations or over-all accu¬ 
mulations. 

The Computer operates at a basic speed of 150 cards per minute but 
will not proceed to the next card until the computation of the card then 
being processed is completed. In other words, for most commerical prob¬ 
lems the Computer maintains its basic speed of 150 cards per minute, but 
on long iterative problems the output speed will be lowered in order that 
the Computer may complete the computation. 

The Computer has provision for 40 program steps. The programming is 
accomplished by means of a flexible wiring panel. 

Features of the Remington Rand Punched Card Computer are: (1) each 
program step is self-checking without requiring another program step; (2) 
the program steps may be used as often as required by means of “branch¬ 
ing” or “selectors”; (3) alphabetical data may be transferred from “mas¬ 
ter” cards to following “detail” cards; (4) the dual card receiving magazine 
permits the segregation of “master” cards from “detail” cards or negative 
from positive results, etc. 

The Computer is also capable of solving problems in higher mathematics 
thereby making it suitable for engineering, scientific and research programs. 

“Synchro-niatic.” This electrical synchronization of a tabulating 
card punch and a Remington bookkeeping machine automatically punches 
information into tabulating cards simultaneously with the recording of 
that information on an original record by means of the accounting machine. 
The latter controls the rate of the operation. 

The Summary Card Punch (Figure 3-48) punches alphabetical char¬ 
acters and numerical information, both designating and adding, into a 




COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 73 



Fig. 3-48. Remington Rand summary card punch, combined with an alphabetic 
tabulator. 

tabulating card automatically and simultaneously with the printing of 
the information as a group total by the tabulator. The capacity of this 
machine is determined by the number of group totals produced by the 
tabulator with which it is connected. 

Tag Control Reproducer is a tabulating card punch which reads “pin 
holes” punched in garment price tags and in turn, at a speed of 100 tags 
per minute, punches the recorded data such as vendor, style, size, color, 
selling period (season), store number or department, price and cost in 
standard tabulating cards. 

These cards are then used to prepare reports of sales by the various codes 
mentioned above, as well as inventory balance reports, thereby providing 
management with up-to-the-minute reports for effective merchandising. 

Tape-to-Card (Figure 3-49) and Card-to-Tape Converters are also 
being used in industry and transportation. Data typed on communication 
equipment such as “teletype” is also recorded by means of holes punched 
in paper tape. This tape in turn activates the tape-to-card unit which 
punches the data in tabulating cards. 

Similarly, data recorded in punched cards, which are then put through 



74 


PUNCHED CARDS 



Figure 3-49. Remington Rand tape-to-card converter. 


the Card-to-Tape Converter, results in a punched paper tape which may 
then be transmitted automatically over wire communication systems to 
remote points. This equipment was originally developed for the transpor¬ 
tation industry, but industry in general has found many uses for the same 
machines and is presently using it to tie in remote warehouses with central 
accounting and billing points. 

SUPPLEMENTS TO PUNCHED CARDS 
Filmsort 

Filmsobt Division, Dextek Foldeh Company 
Pearl River , N. Y. 

The Filmsort system converts microfilm into a punched-card tool. 
Reference and research materials of varying length and size are copied on 
16- or 35-mm microfilm with standard cameras. The individual strips or 
frames are fitted into standard, uniform Filmsort cards, which may be 
plain index cards, edge-notched punch cards, or standard tabulating cards. 

After microfilm insertion, the punched cards may be processed through 




76 


PUNCHED CARDS 


insertion by Filmsort. Filmsort takes the cards which the customer has 
selected, die cuts a window and applies a pressure-sensitive adhesive that 
holds the microfilm. A glassine sheet is put into the window to protect the 
adhesive until the microfilm is inserted into place. 

After the film is inserted into the card, the punched card with microfilm 
may be sorted, selected, collated and handled in any of the standard 
punched-card procedures. 




Figure 3-51. Filmsort jacket cards, paper and acetate. 












COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 77 


Filmsort Jacket Card (Figure 3-51). Developed for the multi-paged 
record of varying length, the Filmsort jacket card condenses a file folder 
of data to standard card size. Strips of microfilm are inserted into the 
slot between the two layers of the jacket and the microfilm can be read 
without removing it from its jacket. 

Jackets are manufactured in acetate.or paper models, in standard 3 x 5, 
4 x 6, and 5x8 inch sizes. In cooperation with Royal McBee, any Filmsort 
paper jacket can be processed for McBee form printing and Keysort 
marginal punching techniques. 

Jacket cards condense the contents of 12 to 14 file cabinets into one 
cabinet of Filmsort cards. The acetate jacket stores the most film in the 
least space. A report of 120 pages, each 8Y x 11 inches, becomes as small 
as one Filmsort 5x8 inch jacket. 

Filmsort Mounter (Figure 3-52). More than 350 individual microfilm 
frames may be mounted hourly into Filmsort aperture cards. In one stroke, 
the mounter cuts the frame of the film from the microfilm reel and simul¬ 
taneously attaches it to the Filmsort card. The rate of production for the 
mounter compares favorably with that of standard planetary cameras. A 
new automatic machine (Figure 3-53) for mounting film into cards at the 
rate of 2,000 cards per hour has recently been developed. 

Filmsnips (Figure 3-54). This tool is a hand-operated scissors-like die 
used to cut out microfilm for manual insertion into Filmsort aperture 
cards. It is recommended for those who mount 50 frames of film or less 
per day. 



Figure 3-52. Filmsort mounter. 




78 


PUNCHED CARDS 



Figure 3-53. Filmsort automatic mounter. 



Figure 3-54. Filmsnips. 


Filmsort Inspector (Figure 3-55). This compact table top viewer has 
an 11 x 11 inch screen. It weighs less than 20 pounds and is usable in all 
microfilm installations where copy is in the x U and 9 x 14 inch range. 
It is equipped for viewing both apertures and jackets and comes in two 
different sizes for 16 or 22 times magnification. 

Filmsort Surveyor (Figure 3-56). Built for viewing large-sized copy, 
the Filmsort Surveyor comes in two models—18 x 24 or 24 x 36 inch screen 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 79 



Figure 3-55. Filmsort Inspector. 


size. It is equipped with variable magnification and has an enlargement 
ratio that can be adjusted from 10.5 to 22 times. 

Magnification is automatic by throwing a switch which starts a chain 
drive that increases or decreases the image. The Surveyor covers a full 
1 1 4 x 1% inch microfilm frame. 

Filmsort Reviewer (Figure 3-57). Where a high intensity light source 
is needed for viewing microfilm, the Filmsort Reviewer is applicable. Its 
optical system and reflectionized screen give faithful reproduction of such 
complicated microfilmed materials as radiographs, charts, graphs, etc. 
Because of its high light source, the Reviewer can be employed as a micro¬ 
film enlarger, using silver chloride papers. 

Reading and Enlarging. All film is read or enlarged directly from the 
Filmsort aperture or jacket. Most standard microfilm enlargers will take 
any Filmsort jacket and the majority of Filmsort aperture cards. 

Reproducing Aperture Cards. Two methods have been developed for 
making positive microfilm duplicates from an original Filmsort aperture 

































COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 81 



Figure 3-58. Filmsort card-to-card duplicator. 


card. The Kalfax process uses an ultraviolet-sensitive film which is devel¬ 
oped and fixed by heat. The Ozalid process uses a card with an unexposed 
ozalid duplicating microfilm in the aperture. It is processed by the standard 
diazo process. 

A recent development is the Card-to-Card Duplicator (Figure 3-58) 
which is capable of reproducing Filmsort cards complete with mounted 
image at a rate of 2,000 an hour. 

Distribution. Remington Rand, the Recordak Corporation, the Ozalid 
Division of General Analine and Film Corp., and Microdealers, Inc. are 
the national distributors for the various Filmsort products. 

Microtape 

The Microcard Corporation 
West Salem, Wis. 

Microtape (Figure 3-59) was developed by The Microcard Corporation 
as an economically feasible substitute for Microcards, where less than five 
copies of a document were needed. Microtape consists of 100-foot rolls of 
16-mm or 35-rnrn positive microtext, printed from standard negative micro¬ 
film rolls, and having a pressure-sensitive adhesive laminated to the back. 




82 


PUNCHED CARDS 



Figure 3-59. Microtape, in reel form and applied to card. 


The user cuts apart the strips, peels off the protective backing, and presses 
the Microtape onto the filing card. Microtape can be applied to standard 
index cards, visible-index cards, and edge-notched cards. It cannot be 
used with tabulating punched cards at the moment because of its added 
thickness. 

Three readers are available, two of which are desk models. One of these 
has a card-moving mechanism for ease in locating pages. The third is a 
portable pocket model with a battery and 110 volt sources of illumination. 

The American Microfilming Service Company, Microtape Systems, of 
New Haven, Connecticut, has set up, under license from The Microcard 
Corporation, a chain of processors who will either arrange for the produc¬ 
tion of Microtapes from existing microfilms or take the microfilms and 
have the Microtapes manufactured. 

OTHER TYPES OF INFORMATION SEARCHING DEVICES 
Uniterm, Matrex and Radex System 

Documentation, Inc. 

2521 Connecticut Ave. NW ., Washington 8, D. C. 

The Uniterm System of coordinate indexing is a manual information 
retrieval system. It provides a card for each indexing term used to charac¬ 
terize the documents in a collection. 

Each of these Uniterm cards is divided into columns headed by the 
digits “0” through “9”. The number of a report to be indexed by a given 
term is entered on the term card in one of ten columns. The digit at the 
head of the column corresponds to the last digit of the report number. The 
term cards are filed alphabetically. 

The collection is searched by comparing the term cards corresponding to 





Figure 3-61. Uniterm book system. 


COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 83 


Figure 3-60. Uniterm card system. 

the search question, for common numbers. These constitute the search 
results. An application of the Uniterm System is described in Chapter 7. 

The Uniterm System comes in two forms. The Uniterm-Card system, 
illustrated in Figure 3-60, is used where only a single index is required. 
Uniterm cards and instruction books are available from Documentation, 
Inc. 

Where the need exists for a low-cost large scale dissemination of an in¬ 
dex, the Uniterm-Book system illustrated in Figure 3-61 can be used. Each 







84 


PUNCHED CARDS 


page carries the reproduction of a number of Uniterm cards. In order to 
facilitate the comparison of terms for common numbers, two duplicate 
sets are provided within a single binding. This eliminates the need to turn 
pages in making comparisons. Uniterm-Book systems are prepared by 
Documentation, Inc. on a contract service basis. Information for Industry, 
Inc., Washington, D. C., publishes a current index to Chemical Patents 
based on the Uniterm-Book principles. 

The Matrex (matrix-index) systems are machine systems specifically 
designed for information retrieval. Two Matrex systems now available are 
the Termatrex and Alpha-Matrex systems. They are based upon cards 
each representing the entire collection in the form of a matrix, as first 
proposed by Batten. 

Termatrex systems feature a card for each indexing term used to 
characterize the item of information in a collection. Each term card has a 
small area dedicated to each document in the collection. 

An item of information is entered into the Termatrex system by placing 
all term cards corresponding to the indexing terms by which that docu¬ 
ment has been indexed, in superimposition in the Termatrex device. A 
hole is then “punched” with a high-speed drill in all of these cards simul¬ 
taneously. 

The collection is searched by placing the term cards corresponding to the 
search question in superimposition in the same Termatrex device. The 
search results, in the form of coinciding holes in the question cards, are 
visible as light dots on a screen. The serial numbers of these items of in¬ 
formation are read off directly. 



Figure 3-62. Termatrex-15 device. 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 85 



Figure 3-63. Termatrex-10 device. 


Termatrex devices come in three models: 

Termatrex-15, shown in Figure 3-62, has a capacity of 15,000 items of 
information per set of 5 x 8 inch cards. It is intended for private collections 
where low cost is a first requirement. 

Termatrex-10, shown in Figure 3-63, has a capacity of 10,000 items of 
information per set of 10 x 10 inch cards. It features wider spacing and 
larger holes, and is used for smaller library and industrial applications 
where ease of operation is a first requirement. 

Termatrex-40, shown in Figure 3-64, is the all-purpose standard Matrex 
machine. It has a capacity of 40,000 items of information per set of 17^ x 
17inch cards. 

For applications where the data input load or the search load is extremely 
high, Alpha-Matrex systems featuring a higher degree of mechanization, 
are available. 

In the Alpha-Matrex systems, cards are provided for each letter of the 
alphabet rather than for each indexing term. This results in a relatively 
small and fixed number of cards, and in a high storage efficiency. From 
seven to ten alphabets are used to avoid spelling ambiguity, (first letter 
alphabet, second letter alphabet, etc.). 

In entering a new item of information in the system, the indexing terms 
used to index the item are typed out on a keyboard which actuates the 
Alpha-Matrex selector to deliver the corresponding cards. The selected 
cards are placed in the Termatrex-40 and the item is entered by “punch- 




8G 


PUNCHED CARDS 



Figure 3-64. Termatrex-40 device. 


ing” a hole in all cards simultaneously at the position corresponding to the 
serial number of the item. 

For searching, a similar selection procedure is followed. The selected 
cards are placed in the Termatrex-40, and the serial numbers of light dots 
visible on the display screen are read out. 

The Alpha-Matrex comes in two models. The mechanical Alpha-Matrex, 
shown in Figure 3-Go, features an adding-machine type of keyboard. 

The electrical Alpha-Matrex system, shown in Figure 3-66, features an 
electric typewriter which also provides a record of the typed terms for 
verification purposes. Errors can be corrected by means of correction but¬ 
tons on the selector. After all terms have been typed out, the corresponding 
cards are instantaneously selected. These are then placed in the Terma¬ 
trex-40 machine. The capacity of the Alpha-Matrex systems is 40,000 items 
per set of cards. 

Matrex systems can be used with any type of indexing system, whether 
it be based on Uniterms, Descriptors, a Subject Heading system or a 
classification schedule. Existing collections based on a Subject Heading or 
classification schedule can be entered into a Matrex System without re¬ 
indexing. This form of mechanization combines the retrieval possibilities of 
the Subject Heading or classification index with the retrieval possibilities 
provided by Uniterm or Descriptor indexing. 



COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 87 



Figure 3-65. Mechanical Alpha-Matrex device. 


Matrex systems can be provided with simple photographic equipment 
for photo-logical operations. This equipment allows a complete program 
comprising logical sums, products and complements, as well as generic 
searches, to be performed on the entire information collection in one se¬ 
quence and in a matter of minutes. 

The Mega-Matrex system, which is built to customer specifications, is 
capable of handling collections running into millions of items. 

The Radex filing system is a novel numerical filing system, applicable 
to cards 5x8 inches and larger. The cards are grouped in decks of 100 and 
each card is provided with a small tab on which its terminal digit has been 
printed. There are a hundred different tab positions corresponding to each 
of the hundred different cards contained in the deck. 

This system allows instantaneous identification of each card within a 
deck of 100. It likewise allows random refiling within each deck. Figure 
3-63 shows Termatrex cards equipped with Radex tabs. 




88 


PUNCHED CARDS 



The Filmorex System 

74, Rue des Saints-Peres 
Paris ?*, France 

The Filmorex system operates with pieces of film (Figure 3-67) coded 
by means of light and dark spots, in much the same way that tabulating 
cards are internally punched. Each piece of film, 70 x 45 millimeters, is 
divided into two sections. One is a microphoto of the document and the 
other is the light and dark pattern of appropriate codes used for searching. 

A microfilm camera is available with provision for producing the proper 
code patterns by means of a keyboard. The other main piece of equipment 
is the selector (Figure 3-68), which will read and select film at the rate of 
36,000 pieces per hour. 

Magnetic Inks, Punched Paper Tape, Magnetic Tape, Etc. 

In the August, 1956, issue of Banking there appeared a report of a sub¬ 
committee of the American Bankers Association which investigated the 
mechanization of check handling by means of magnetic and luminous inks. 
About ten companies are listed there as participating in the research and 
development of these methods. 

The June, 1957, issue of Computers and Automation contains an ex- 




















COMMERCIALLY AVAILABLE EQUIPMENT AND SUPPLIES 89 



Figure 3-67. Filmorex film. 



Figure 3-68. Filmorex selector 














90 


PUNCHED CARDS 


haustive list of manufacturers of computers, components, magnetic and 
paper tape readers and writers, and other information processing equip¬ 
ment. 

An article appeared in Science (October 26, 1956) by Karl Heumann on 
the use of computers and related devices in the field of information re¬ 
trieval. The reader is referred to the above mentioned sources for further 
information on non-punched card equipment in use, or for possible use, in 
the storage and retrieval of information. 

ANCILLARY EQUIPMENT 

Many companies manufacture ancillary equipment such as card files, 
key punch desks, plugboard storage racks, index cards, and the like. Some of 
them are listed below. Again, this is not to be considered an exhaustive 
list. 


Art Steel Company, Inc. 

Monarch Metal Products, Inc. 

170 West 233rd St. 

724 South Columbus Ave. 

New York 63, N. Y. 

Mount Vernon, N. Y. 

Berger Division 

Record Files, Inc. 

Republic Steel Corporation 

1490 Lincoln Highway West 

E. 11th Street & Belden Ave. 
Canton 5, Ohio 

Wooster, Ohio 


Shaw-Walker 

Duro Consolidated, Inc. 

1950 Townsend St. 

P. 0. Box 248-1 

Redwood City, Calif. 

Muskegon 6, Mich. 


Tab Products Company 

Dresser Products, Inc. 

57 Post St. 

152 Wheeler Ave. 

San Francisco 4, Calif. 

Providence 5, R. I. 

The Wright Line, Inc. 

Globe-Wernicke Company 

100 Exchange St. 

5029 Carthage Ave. 

Cincinnati, Ohio 

Worcester 8, Mass. 

Acknowledgment 



Mr. Harry C. Zeisig, Jr., of the Spencer Chemical Company, originally assembled 
much of the information for this chapter. Due to the pressure of other business he 
was unable to complete the collection of material. The author wishes to acknowledge 
Mr. Zeisig’s important contribution to this chapter. 



Part II 

PRACTICAL APPLICATIONS OF 
PUNCHED CARDS AND 
RELATED DEVICES 




Chapter 4 


THE USE OF PUNCHED-CARD TECHNIQUES 
IN PREPARING REPORTS, PAPERS, 

AND BOOKS* 


Charles A. Burkhard 

Locomotive and Car Equipment Department, General Electric Co., Erie, Pennsylvania 

Introduction 

A research project is not finished until the results have been reduced to 
writing, in the form of a report, a paper, or even a book. Report writing 
and manuscript preparation have been widely discussed. This chapter will 
deal only with the mechanical aspects of the problem. 

In general, an outline is prepared and pertinent data are then inserted in 
their logical places. This step involves selecting, sorting, arranging, and 
correlating large masses of data. Other chores connected with technical 
writing, such as the arrangement of the bibliography, also involve much 
tedious shuffling of papers or cards. If conventional methods are used, these 
tasks, the performing of purely mechanical operations, may take an undue 
proportion of the writer’s time. This chapter will show how punched-card 
techniques can be used as an aid to technical writing by facilitating the 
time-consuming mechanical operations. To illustrate the effectiveness of 
such techniques an account is given of the use of punched cards in preparing 
a review paper on organosilicon chemistry. 1 

Description of Punched-Card Filef 

Author, title, reference, and abstract, including physical and chemical 
data, as well as other pertinent bits of information, were written or typed 
on the cards. In general, all information of interest found in each paper, 

* This chapter is based on a paper, entitled “The Use of Hand-Sorted Cards in 
Small Files”, presented before the Division of Chemical Education of the American 
Chemical Society at the 113th national meeting in Chicago, Ill., April, 1948. Thanks 
are due the American Chemical Society for approving publication of the revised 
paper in this book. The author wishes to acknowledge the valuable help given by 
Dr. A. E. Newkirk, Research Laboratory, General Electric Co., Schenectady, N. Y., 
and by others. 

1 Burkhard, Rochow, Booth, and Hartt, Chem. Rev., 41, 97 (1947). 

f Five by eight-inch McBee “Keysort” cards were used (Figure 4-1). 


93 



94 


PUNCHED CARDS 


publication, or company report was entered on a single card. The coding 
and punching of each card was arranged to cover completely the information 
entered on the card. 

The coding scheme is based on the following general headings: 

A. General subject 

B. Compounds described 

C. Author 

D. Date of publication 

E. Card serial number 

A. General Subject Code. Space for direct coding of 32 general sub¬ 
jects is provided by 16 double holes along the upper edge of the card; shal¬ 
low and deep punching are used. By using the intermediate punch, also, it 
would be possible to code a total of 48 subjects. 

Numbers 1-16 are indicated by shallow punching, and numbers 17-32 
by deep punching. The subject code was arranged so as to minimize over¬ 
lapping of shallow and deep punching. Inconvenience due to overlapping 
has been negligible. 

Table 4-1 gives some typical entries in the subject code. Since the sub¬ 
jects are direct coded, it is possible to code on each card all the subjects 
covered in the reference listed on that card. 

B. Code for Compounds Described. This code is based on the var¬ 
ious types of bonds and structural units found in organosilicon compounds. 
The code has been set up as a decimal classification. Nine important func- 

Table 4-1. Subject Code 

1. Analytical method (a new method) 

2. Bibliography 

3. Book 


10. Patent 

11. Nomenclature discussion 


14. Resins 

15. Review articles 


17. Structure investigation 



TECHNIQUES IN PREPARING REPORTS, PAPERS, AND BOOKS 95 


Table 4-2. Chemical Code 


1.00 

Si-H 

2.00 

Si-Halogen (general) 


2.10 Si-F 


2.20 Si-Cl 


etc. 

3.00 

C-Halogen 

4.00 

Si-OH 

5.00 

Si-C 


5.10 Si-Alkyl 


5.11 Si-CHj 


5.12 Si-C,H, 


5.20 Si-Aryl 


5.21 Si-C,H, 


etc. 

6.00 

Si-Si 

7.00 

Si-O-Si 

8.00 

Si-O-C 

9.00 

Si-Miscellaneous 


tional or characteristic structural features were chosen as the main headings 
in the code. These nine headings were then further divided and subdivided. 
The general procedure is illustrated in Table 4-2. 

A chemical compound is coded on a card by punching the appropriate 
classification numbers in the outer row of holes along the bottom of the 
card, in the section marked “Compound Index”. For example, to code the 
class of compounds, Si-C«H # , its number, 5.21, is punched as 5 in the units 
field, 2 in the tenths field, and 1 in the hundredths field. 

If more than one compound is mentioned in the reference, each such 
compound is coded in this same field. All cards bearing information about 
a given compound or group of compounds may then be mechanically se¬ 
lected from the file by procedures described in Chapters 2 and 3. Occa¬ 
sionally such a sort selects from the file not only all the cards desired but 
some extra ones, due to the superimposed coding. It has been found in 
actual practice that such extra cards are relatively few in number and are 
readily eliminated by inspection. 

A mathematical analysis of superimposed coding is given in Chapter 21. 

C. Author Code. If there is only one author, the first three letters of 
his surname are coded. If there are two or more authors, the first letters 
of the surnames of the first two and last authors are coded. The hole marked 
II is also punched to indicate plural authorship. 

The holes assigned to coding author names are along the left edge of the 
card. The author code is based on the OIECB mnemonic code described 
by Casey, Bailey, and Cox 2 . By applying the serial sorting procedure, de- 

* Casey, R. S., C. F. Bailey, and G. J. Cox, J. Chan. Ed., 23, 496-9 (1946). 



96 


PUNCHED CARDS 


scribed in Chapter 2, the entire file may be arranged alphabetically by au¬ 
thors in a very short time. Some hand-sorting is required with multiple 
author cards. 

Good results have also been obtained when this method of coding was 
used for locating a multi-author reference, the known-author’s name of 
which was not that of the first author. In an actual case, it was desired to 
locate a reference known to bear a certain well-known author’s name. After 
about ten-minutes searching the desired reference was located, and it was 
found that the known-author’s name was the second among several other 
authors. 

D. Date of Publication. The year of publication is coded in the holes 
located along the upper edge of the card and at the right of those used for 
the General Subject Code. The century of publication is directly coded in 
holes marked 17, 18, 19, while the decade and year are numerically coded 
using the 7-4-2-1 scheme described in Chapter 2. 

E. Card Serial Number. Each card is assigned a serial number. This 
number is written on the card in the space marked I, and coded in the 
fields at the left edge of the card, as shown in Figure 4-1. Such a numbering 
system is extremely useful. The serial number may be used to identify the 
reference, as discussed in more detail below. Also, the last serial number 
assigned indicates the total number of cards in the file. 

Use of the Punched Card File in Writing a Review Paper 


The most important use of the file in writing the review paper was in 
selecting information pertinent to the subject matter discussed under the 



Figure 4-1. Card used for information on organosilicon compounds. 



















TECHNIQUES IN PREPARING REPORTS, PAPERS, AND BOOKS 97 


various headings and subheadings of the outline: 

I Introduction 

II. Nomenclature 

III. Methods for the synthesis of organosilicon compounds 

A. Synthesis of organosilane and organochlorosilanes 

B. Synthesis of organofluorosilanes 

1. By the Grignard reaction 

2. From silicones 

3. From organochlorosilanes 

C. Synthesis of organochlorofluorosilanes 

IV. Behavior of classes of organosilicon compounds 

A. Normal alkyls of the type SiR 4 , RiSiSiRi , etc. 

B. The alkylsilanes R„SiH 4 - n 

C. Organosilanes with substituted alkyl and aryl groups 

D. Alkyl and aryl halogenesilanes, R n SiX 4 _ n 

1. Properties of organochlorosilanes 

2. Properties of organofluorochlorosilanes 

E. The alkylalkyloxy- and aroxy-silanes R n Si(OR) 4 _ n 

F. Organosilanols RnS^OHVn , etc. 

G. Esters, CH*COOSiR, and (R,SiO)*SO, 

H. The linear and cyclic organosiloxanes 

I. Silazanes and related groups 

V. The silicone polymers 

VI. Water-Repellent films 

VII. Special investigations of physical properties 

VIII. Isomerism 

IX. Physiological properties 

X. Analytical methods 

XI. References 

The cards bearing references to the desired subjects or types of com¬ 
pounds were selected from the punched-card file by sorting at the appro¬ 
priate positions (holes). For example, before the section on Nomenclature 
was written, the cards were sorted in position number 11 in the subject in¬ 
dex. In this way all the references pertaining to this topic were dropped 
from the file. It was then quite simple to draft this section directly from 
the cards. Likewise, the other sections of the publication were drafted by 
sorting first for the subject under consideration and subsequently rear¬ 
ranging the selected cards, by sorting procedures described in Chapter 2, 
into chronological order or according to subtopic to fit into the outline that 
had been previously prepared. For example, under Section V, entitled, “The 
Silicone Polymers”, various physical properties of this class of compounds 
are dealt with. It was possible to separate first the silicone oil and rubber 
and resin topics from the file, and then arrange these so that the topics 
pertaining to viscosity, molecular weight, oil properties, resin properties, 
and rubber properties could be dealt with in their logical order. 

The review paper also included a number of tables giving physical data 



98 


PUNCHED CARDS 


for various types of organosilicon compounds. In preparing the tables of 
organosilicon compounds the cards were sorted, using the Compound Index, 
to select the particular class of compounds desired. Each compound was 
then transcribed from the card together with its physical properties and 
card number. As a typical example, one of the tables contains the alkoxy- 
and aroxyorganosilanes. It was possible by sorting for 8 in the Compound 
Index to separate from the file those cards that had data pertaining to com¬ 
pounds containing the ether group. Then, by hand-sorting, these were ar¬ 
ranged in the order desired. 

Preparing the Bibliography 

The punched-card file proved very helpful in preparing the bibliography 
of the review paper. During the preparation of the manuscript the card 
serial numbers were used as provisional reference numbers. When the final 
draft of the manuscript had been completed, the cards which had been 
used in its preparation were removed from the file and arranged alpha¬ 
betically by author by serial-sorting the author index. The bibliography was 
then prepared directly from the cards by typing the references in alpha¬ 
betical order by author. These references were numbered in sequence, and 
each such number was penciled on the corresponding card. To facilitate 
placing these final bibliography numbers in their proper places in the manu¬ 
script the cards were then arranged in order according to card serial number 
by sorting the card number index. It was then easy to scan the manuscript, 
find the card corresponding to each serial number used as temporary ref¬ 
erence number, and replace the latter with the final bibliography number 
penciled on the card. 

Advantages of Using Punched-Card Files for Preparing Papers* 

Reports, etc. 

The application of punched-card techniques permitted the use of rapid 
mechanical methods to avoid tedious and time consuming hand-sorting 
and individual scanning of papers and cards. 

The file on organosilicon compounds has also proved very useful in seek¬ 
ing and establishing correlations, such as cause-and-effect relationships. 
This is a matter of very great importance in preparing reports and papers 
concerned with previously unpublished data. 

While discussing the advantages of punched cards for filing scientific and 
technical information, it should not be forgotten that the same file used in 
writing final reports, papers, etc., can also be used to good advantage in 
planning and executing experimental work. The author’s file on organo¬ 
silicon compounds was often consulted by himself and colleagues at the Gen¬ 
eral Electric Co. when experimental work in that field was planned. 



TECHNIQUES IN PREPARING REPORTS, PAPERS, AND BOOKS 99 


Briefer papers than the one referred to above have also been prepared 
using the same file. Thus, for a special report it was required that all the 
data pertaining to organosilicon compounds that contained the Si—OH 
group be obtained. The accompanying bibliography was to be placed in 
chronological sequence. The punched-card file was first sorted in the Com¬ 
pound Code field for 4 the code number for Si—OH. The cards isolated by 
this sorting operation were then arranged in chronological order by serial 
sorting the date of publication index (upper right field in Figure 4-1). The 
total time required for manipulating the punched cards was about 15 min¬ 
utes. 

In addition to entering numerical data and similar reference information 
on punched cards it is also possible to attach photographs, drawings, graphs, 
charts, microfilms, etc. directly to the cards. By sorting for a particular 
item or topic all corresponding photographs, drawings, and charts may be 
obtained in addition to the pertinent references and data. This is often 
helpful in writing papers as it may prove desirable to include photographs 
etc., which otherwise might be overlooked. 

Keeping track of abstract references when setting up a file is facilitated 
by the following procedure: When searching abstract and journal indexes 
for a given compound or subject it is helpful to establish a special card— 
termed “Abstract Card”—on which all abstract references are listed to¬ 
gether with any index information concerning the abstract. The compound 
or subject is edge-punched in the card in the usual way. When an original 
paper is read and a card prepared for that reference, the corresponding 
abstract notation on the Abstract Card is crossed with a light pencil mark. 
In this way the Abstract Card serves as a record of the search performed on 
a given subject or compound. A special hole is reserved to designate the 
Abstract Card and enables one to separate the Abstract Cards from the 
other cards in the file. 

The hole designated by III in Figure 4-1 is used in the file just described 
to designate a company report. 



Chapter 5 


AN INTERNATIONAL CLASSIFICATION 
AND PUNCHED-CARD FILING SYSTEM 
FOR METALLURGICAL LITERATURE 


Marjorie R. Hyslop 

Editor of A. S. M. Review of Metal Literature, American Society for Metals 

Cleveland, Ohio 

AND 


Alvina Wassenberg 

Research Librarian, Division of Metallurgical Research 
Kaiser Aluminum & Chemical Corporation, Spokane, Washington 

Literature searching has long been accepted as a preliminary and highly 
important step in any research problem. It is tedious and time-consuming 
work. The excellent facilities offered by the various public, private and 
technical libraries to smooth the way of the researcher cannot be mini¬ 
mized. Nevertheless, a large part of the burden still remains the responsi¬ 
bility of the individual investigator. 

Two problems face the technical researcher or investigator who must 
make up a file of technical information. First is a logical and usable analysis 
of the subject matter in his particular field of interest, either in the form of 
a list of subject headings or a classification outline. Second is the method 
of handling his files of information or literature references and the tools 
used therefor. 

Recognition of this problem in the field of metallurgy was evidenced as 
early as 1947 by Guy and Geisler. 1 Response to the publication of this ar¬ 
ticle was so enthusiastic that the eventual result was the formation of a 
joint committee by the American Society for Metals and the Special 
Libraries Association, whose function was to design a standardized classi¬ 
fication system for metallurgical literature and a punched card filing system 
for use with it. The committee work was completed early in 1950 with the 
publication of the “ASM-SLA Metallurgical Literature Classification,” 2 - * 

1 Guy, A. G., and Geisler, A. H., “A Punch Card Filing System for Metallurgical 
Literature,” Metal Progr., 62, 993 (Dec. 1947). 

* ASM-SLA Metallurgical Literature Classification, prepared by a Joint Com¬ 
mittee of the American Society for Metals and Special Libraries Association; Ameri¬ 
can Society for Metals, Cleveland, Ohio, 1950. 

* Geisler, A. H., “How to Find Detailed Information When You Want It,” Metal 
Progr., 67, 613 (May 1950). 


100 



CLASSIFICATION FOR METALLURGICAL LITERATURE 


101 


and the underwriting of the manufacturing charges for a large supply of 
punched cards especially printed to accommodate the provisions of the 
classification. 

In the intervening seven years the classification has been widely adopted, 
and the first printing of 2000 copies was exhausted early in 1956. It has 
been found during this time that the classification lends itself very well to 
organization of literature files that are not necessarily based on punched 
cards. To facilitate the organization of such files many abstracting services 
and technical journals, both in this country and abroad, print the classi¬ 
fication code symbols in conjunction with titles of articles or abstracts. In 
this country all of the articles published in Metal Progress , Transactions of 
the American Society for Metals and all of the abstracts in the AJ5.M. Re¬ 
view of Metal Literature are so coded. 

In Italy, the Associazione Italiana di Metallurgia (Italian Association 
of Metallurgy), the Istituto Siderurgico Finsider (ferrous metallurgy) and 
the Institute Sperimentale dei Metalli Leggeri (light metals) have adopted 
the ASM-SLA system. 4 - 6 These three organizations code their abstracting 
services and also the principal articles in their official journals with the 
ASM-SLA symbols, similar to the practice of a number of European pub¬ 
lishers who print the symbols, of the Universal Decimal Classification 
(U.D.C.). 

In England, the classification and punched card system have been 
adopted by the Information Section of the British Iron and Steel Research 
Association,® and in Germany by the Institut fur Harterei-Technik, which 
also codes the principal articles and the abstracts published in its journal 
Hdrterei-Technik und W&rmebehandlung. 

A new joint committee of the American Society for Metals and the Spe¬ 
cial Libraries Association was organized in 1955 to revise the classification 
and bring it up to date for a second printing (See Appendix A). 7 The pur¬ 
pose of this revision is to provide for new fields of scientific knowledge (such 
as metallurgical aspects of atomic energy), as well as to expand some of 

4 Classificazione Bibliografica Internazionale della Metallurgia, First Italian 
Edition of the ASM-SLA Classification; Italian Association of Metallurgy, Milan, 
Italy, 1955. 

* Documentazione Tecnica a cura del Centro Documentazione A.I.M.; La Me¬ 
tallurgia Italiana, 47, No. 6, 186 (June 1955). 

* Colinese, P. E., Why B.I.S.R.A. Has Adopted the American Society for Metals— 
Special Libraries Association (ASM-SLA) Metallurgical Literature Classification; 
Aslib Proceedings, 5, No. 4 , 345 (Nov. 1953); Metals Review, 27, No. 12, 26 (Dec. 
1953). 

7 Committee Formed to Study ASM-SLA Literature Revision; Metals Review, 
Vol. 28, No. 7, July 1955, p. 14; 30, No. 26 (Feb. 1957); see also L. S. Foster, “Revi¬ 
sion of the ASM-SLA Code for Metallurgical Literature." Paper presented before 
the Division of Chemical Literature, 131st National Meeting of the American Chem¬ 
ical Society, Miami, Florida, April 8, 1957. 



102 


PUNCHED CARDS 


the sections in somewhat greater detail. Additions and revisions are made 
in such a manner that the general outline of the first edition can be re¬ 
tained so that existing files are not invalidated. A parallel committee of the 
Italian Association of Metallurgy is working closely with the American 
committee in this revision so that the new edition of the classification 
will represent a joint undertaking which can eventually be adopted as an 
international standard for metallurgical literature classification. 

While there are no limitations on the size of literature files that can be 
maintained using the classification code symbols, the limitations for a 
punched card file are somewhat narrower. Sales of the punched card, which 
was especially designed for use with the system, have continued at a steady 
rate; by early 1956 sales had reached a total of more than two million 
cards. While it is difficult to estimate the number of individual files repre¬ 
sented by this total, a conservative guess would be in the neighborhood of 
several hundred at least. 

The upper limit for hand sorting marginal punched card files is estimated 
to be about 10,000 literature references before the system breaks down be¬ 
cause of the limitations of hand sorting and needling. The punched card 
system is therefore primarily suited to the needs of the individual researcher 
whose files are more likely to number somewhere between five hundred and 
a couple of thousand references. It is also proving useful to the technical 
librarian whose interests are largely confined to some aspect of metallurgi¬ 
cal science and technology, and an installation in a company library will be 
described in some detail later in this chapter. 8 - * A system maintained by a 
plant metallurgist was described before the Metals Division of the Special 
Libraries Association by E. C. Wallace in October 1955. 10 

The Metallurgical Literature Classification 

Previous to the work of the ASM-SLA committee, no thorough analysis 
of the subject of metallurgy as an entity was available. Existing classifica¬ 
tion systems, such as the Dewey Decimal, Universal Decimal, and various 
library cataloging systems, were designed to accommodate all fields of 
science and technology, with the result that metallurgical interests were 
scattered somewhat promiscuously among a variety of other subjects. 
What attempts had been made to treat metallurgy as a separate science 
were either inadequate or impractical. 

No one method of breaking down the subject of metallurgy can be uni- 

* Wassenberg, Alvina, Experience With the ASM-SLA Classification of Metallur¬ 
gical Literature. Paper presented before Metals Section, S-T Division, Special 
Libraries Association, Detroit, Mich., October 1951. 

* Edelman, David L., “Library Use of New Indexing System,” Metal Progr., 59, 
526 (April 1951). 

10 Wallace, E. C., “Use of the ASM-SLA Metallurgical Literature Classification 
System by Industrial Metallurgists,” Special Libraries, 47, No. 3, 114 (March 1956). 



CLASSIFICATION FOR METALLURGICAL LITERATURE 


103 


versally useful to every metallurgist and every librarian. One person may 
be primarily interested in metallurgical processes, another in metallic 
properties, a third in specific metals and materials, a fourth in a specific 
metal-product form (such as tubing, sheet, wire), and still a fifth may wish 
to classify his data by equipment used. 

With this in mind, the committee decided to provide three parallel and 
independent classification outlines or indexes, one a “Processes and Proper¬ 
ties Index,” one a “Materials Index,” and the third a so-called “Common 
Variables Index.” A condensed outline of the entire classification is given 
in Appendix A, where changes and additions appearing in the second edition 
are indicated by italics. In addition, the punched card provides for an 
Author Index and an optional Date Index. 

In preparing these indexes, the committee decided to provide a sys¬ 
tematic, logical and practical breakdown by subject without attempting 
to fit the various headings into any preconceived coding system, such as a 
decimal arrangement. The fundamentals of the punched card, however, 
were borne in mind so that the various headings and subheadings could be 
accommodated on the card with an appropriate coding system consisting 
of combinations of letters and numbers and semantic symbols. It will not 
tie necessary to reproduce the complete classification here, but enough will 
be given to illustrate its general outline and arrangement. 

The Processes and Properties Index resolved itself neatly into twenty 
main divisions (termed for convenience “first-order divisions”) as follows: 
A — General Metallurgical 
B — Raw Materials and Ore Preparation 
C — Nonferrous Extraction and Refining 
D — Ferrous Reduction and Refining 
E —Foundry 

F — Primary Mechanical Working 
G — Secondary Mechanical Working 
H — Powder Metallurgy 
J — Heat Treatment 
K — Joining 

L — Cleaning, Coating and Finishing 

M — Metallography, Constitution and Primary Structures 

X — Transformations and Resulting Structures 

P — Physical Properties and Test Methods 

Q — Mechanical Properties and Tests Methods; Deformation 

R — Corrosion 

S — Inspection and Control 

T — Applications of Metals in Equipment 

U — Allied Fields 

V — Materials (Subdivided in Materials Index) 



104 


PUNCHED CARDS 


Table 5-1. Sample of Second and Third-Order Subdivisions in Processes 

and Properties Index 


A — GENERAL METAL¬ 
LURGICAL 

2. History 

3. Education 

4. Statistics and economics 

5. Plant practice 

a. Materials handling 
0. Industrial relations 

7. Health and safety 

8. Secondary metals, scrap and 

waste disposal 

9. Research organisations 

10. Glossaries, definitions, trade 
names, directories 

B — RAW MATERIALS AND 
ORE PREPARATION 

10. Ore deposits and raw materials 

reserves 

11. Sampling (For Assaying, see 

SUs) 

12. Mining 

n. Open pit 

p. Underground 

q. Hydraulicking 

13. Crushing, grinding and siting 

a. Primary crushing 

b. Secondary crushing and 

grinding 

c. Milling 


d. Screening 

e. Wet siting (classification) 

f. Air classification 

14. Concentration and beneficia- 

tion 

g. Gravity 

h. Flotation 

j. Magnetic 

k. Chemical (leaching, etc.) 

m. Agglomeration tabling 

n. Electrostatic 

p. Settling, thickening, filtra¬ 
tion 

15. Roasting and calcining 

16. Sintering and noduliting 

17. Briquetting 

18. Fuels technology* 
n. Solid 

p. Liquid 

q. Gaseous 

19. Refractories technology* 

21. Fluxes and slags 

22. Addition agents 
n. Ferro-alloys 

p. Other metal additions 

q. Ores 

r. Scrap 

s. Nonmetal lies (coke, carbon, 

etc.) 

a. Oxygen 


C - NONFERROUS EX¬ 
TRACTION AND 
REFINING 

21. Smelting 

a. Blast furnace 

b. Converting (to include 

bessemer) 

c. Reverberatory processes 

d. Electric furnace 

22. Distillation 

g. Reduction 

h. Refining 

23. Electrolytic processes 
n. Electrowinning 

p. Electrorefining 

24. Cyanidation 

25. Vacuum refining 

26. Reduction by metals (Thermit 

processes, etc.) 

27. Cementation (copper on iron, 

etc.) 

28. Separation of metals 

g. Parting 

h. Liquation 

29. Amalgamation 

1. Carbonyl reduction 

2. Hydride decomposition 

4. Halide decomposition 

5. Ingot casting 
n. Tapping 

p. Teeming 

q. Continuous casting 


* Limited to fuels and refractories in general. Use of fuels and refractories in a specific metallurgical 
process (melting, heat treating, welding, etc.) should be indexed in the appropriate process section and 
cross-indexed to fuels or refractories in the Common Variables Index. 


Each of these main divisions was further broken down into some ten to 
twenty second-order divisions. When complexity of the subject matter 
required, these second order divisions were further subdivided into third 
orders. An example of the breakdown into second and third orders is 
given for the first three principal divisions of the classification, namely, 
A — General, B — Raw Materials and Ore Preparation, and C — Non- 
ferrous Extraction and Refining. (See Table 5-1). Provision was also made 
on the punched card for fourth orders to take care of the detailed break¬ 
down required by specialists in certain fields; none of these fourth-order 
divisions, however, were provided by the committee. Ample provision was 
made for expansion of the outline in any direction. 

Formulation of the Materials Index presented some special problems. 
While materials in this metallurgical index are confined to metals, the 
metallurgist has various and sometimes contradictory or overlapping ways 
of grouping them. Metallic alloys may be grouped by base metal (aluminum 



CLASSIFICATION FOR METALLURGICAL LITERATURE 


105 


alloys), by composition (aluminum-copper-magnesium alloys), or by spe¬ 
cific properties or uses (heat-resisting alloys, magnetic materials, bearing 
metals). The ferrous alloys (steels) present still further difficulties since 
references may be required to such diverse and overlapping groups as 
openhearth, bessemer or crucible steel, carbon steels, alloy steels, tool 
steels, stainless steels, and cast iron. Provision for indexing by either com¬ 
position or group or both was made, as will be explained later in the sec¬ 
tion on the punched card. 

The Common Variables Index is a miscellaneous collection of factors 
which modify either the Processes and Properties Index or the Materials 
Index, or which refer to the physical characteristics of the publication. 
Such subject headings as equipment, theory, high and low temperature, 
are common to a large proportion of the first- and second- and even third- 
order subdivisions in the Processes and Properties Index. Such “metal 
forms” as castings, forgings, tubing and coated metals might be considered 
as modifying the Materials Index. Physical characteristics of the publica¬ 
tion include type of literature (theory, research, review, plant description), 
form (patent, specification, report, book) and language. 

The Common Variables Index is one of the distinguishing features con¬ 
tributing to the versatility of the punched-card system. In conventional 
card files, its subdivisions can be incorporated as additional subheadings in 
the Processes and Properties and the Materials Indexes, or carried as sepa¬ 
rate main headings. 

For purposes of coding published articles or abstracts, the symbols used 
for each of the three indexes are sufficiently distinctive that they need be 
separated only by commas; if a more pronounced division of the three 
indexes is desired, semicolons or colons may be used to separate the series 
of symbols required for each index. 

One disadvantage of the conventional alphabetical subject index file is 
that many duplicate cards must be made, each one carrying a single entry. 
For example, a reference may be concerned with furnaces for heat treating 
aluminum alloy castings. Indexing entries would be required for “furnaces,” 
“heat treating,” “aluminum,” and “castings.” In the punched card system 
all of these entries can be indexed on one card, and all may be considered 
as primary headings, and not subordinated as subheadings. 

Design of the Punched Card 

The card selected for the ASM-SLA system is a standard 5 x 8-inch card 
manufactured by E-Z Sort Company and marketed by Lee F. Kollie 
Associates, Inc. of Chicago. (Figure 5-1.) A double row of holes is punched 
on all four edges of the card, with a row of six double holes to the inch in¬ 
stead of the more usual four per inch. 



106 


PUNCHED CARDS 



Figure 5-1. Punched card specially designed for use with the ASM-SLA classifica¬ 
tion of metallurgical literature, showing the location of the four principal indexes. 


The Processes and Properties Index occupies the top edge of the card, 
the Materials Index the left side and bottom left edge, the Common Varia¬ 
bles Index the right edge, and the Author Index the bottom right edge. If 
desired, a fifth index by date can be incorporated in a portion of the space 
allotted to the Common Variables Index. Both direct indexing (a single 
hole assigned to a single specific criterion) and indirect indexing (permuta¬ 
tions of two or more holes to designate a single criterion) are used as dic¬ 
tated by the exigencies of the four indexes. 

Processes and Properties Index. The main divisions of this index are 
directly coded by the alphabetical designations A through V. Additional 
divisions coded W, X, Y, Z, AA, BB, CC, DD, and EE are open for future 
expansion. These letters are assigned to the lower or deep row of holes across 
the top of the card, starting at the left. Each of these “first-order” divi¬ 
sions carries a specific meaning: A — General Metallurgical, B — Raw 
Materials and Ore Preparation, etc. Considered alone, it constitutes a di¬ 
rect index. 

Each of these main headings is followed by a series of second-order head¬ 
ings designated by the numerals 1 through 29, assigned to the upper or 
shallow row of holes at the top of the card. Since shallow hole No. 2 in 
Division A (history) has a meaning different from shallow hole No. 2 in 
Division C (hydride decomposition), the combination now constitutes an 
indirect index utilizing combinations of deep and shallow holes to indicate 
a single meaning. Figure 5-2 shows a card with an abstract mounted on it, 





















































CLASSIFICATION FOR METALLURGICAL LITERATURE 


107 


coded and slotted. The reference is coded BIO and A9 in the Processes and 
Properties Index; a deep notch in B and shallow notch in 10 stands for ore 
deposits and raw material reserves (mineral resources of Turkey) and a 
deep notch in A and shallow in 9 stands for research organizations (Govern¬ 
ment Research Institute). 

Third and fourth orders are provided in the Processes and Properties 
Index to give a finer subdivision of the subject matter. This also is an in¬ 
direct index, using the lower case letters “a” through “s” assigned to the 
deep holes at the top right of the card for third orders and the correspond¬ 
ing shallow holes designated 30 through 45 for fourth orders. * 

The corner holes, marked Gi and G 2 , are for “general” subdivisions of 
first- and second-order divisions, respectively. For example, a reference 
coded C-Gi would indicate that it refers to the general subject of “Non- 
ferrous extraction and refining” rather than any specific divisions of this 
general subject. 

In the indirect type of index, such as the Processes and Properties Index, 
the number of criteria that can be simultaneously coded is strictly limited, 
and it is recommended that no more than three first-order divisions be 
coded on any one card. (As many second-order divisions as may be de¬ 
sired can be indexed under one main division.) For this reason, Section V 

* In arranging the classification, the numeral which occupies the same space as 
a letter designation is omitted from the second-order subdivisions under the first- 
order division, since it would be automatically notched out when the first-order 
division is coded. (See Table 5-1.) Thus in Division A, the numbering of the second 
orders starts with No. 2 instead of No. 1; No. 1 is automatically eliminated when the 
letter A is notched. 

Also the ideal arrangment of this coding system would have been to start each 
second-order subdivision under a main division with No. 1, and each third order 
with the letter “a,” but such an arrangement would have resulted in overcrowding 
in the first portion of the code. To avoid such overcrowding, the arrangment of code 
symbols in second, third, and fourth orders is staggered. Staggering in second-order 
subdivisions is roughly by multiples of ten—Division A starts with No. 2, Division 
B starts with 10, Division C with 21, and Division D jumps back to No. 1 again for 
its first designation. After No. 29 is reached in any one main division, the coding 
drops back to No. 1; for example, in Division C, second-order subdivision No. 29 
u ‘amalgamation” is followed by a parallel second-order subdivision No. 1 “carbonyl 
reduction. ” 

Staggering in third orders is in multiples of five letters. In Division C, for example, 
the first third-order subdivision under the first second order “21 Smelting” is 
coded with the letter “a” (“a. Blast furnace”). The first third-order under “22. Dis¬ 
tillation” is “g. Reduction” and the first third-order under “23. Electrolytic proc¬ 
esses” is “n. Electrowinning.” Had third-order subdivisions been provided under 
“24. Cyanidation,” they would have started again with the letter “a”. After the 
letter “s” is reached, the coding goes back again to the beginning of the alphabet. 
(See “B22s. Nommetallic addition agents,” followed by “B22a. Oxygen as an addi¬ 
tion agent.”) 



108 


PUNCHED CARDS 



Figure 5-2. Punched card with abstract mounted, and coded in BIO and A9 in the 
Processes and Properties Index, Fe and ST in the Materials Index, 12-18 in the 
Common Variables Index, and W,E,I in the Author Index. 


on Materials has been added to the Processes and Properties Index. A notch 
in deep hole V at the top of the card merely indicates that the reference is 
coded in the Materials Index and covers various processes and properties in 
a broad and general way. With such a reference as “Manufacture, Proper¬ 
ties and Uses of Aluminum Alloys,” a notch in hole V obviates the neces¬ 
sity of coding such a reference in half a dozen or so first-order divisions in 
the Processes and Properties Index.* 

Materials Index. The Materials Index is arranged in such a way as to 
permit indexing both by composition and by industrial group. The sixteen 
so-called common elements, indexed directly by chemical symbol on the 
upper left side of the card, include the elements which form the basis of 
most of the common metallic alloys. A deep notch is used to indicate the 
base metal of the alloy, and a shallow notch for alloying elements. For exam¬ 
ple, an aluminum-copper-magnesium alloy is notched deep in the hole 
designated A1 and shallow in the holes designated Cu and Mg in Figure 
5-3. 

More detailed breakdowns are provided by utilizing the section at the 
bottom of the card designated third and fourth orders in the Materials 

* In the 1957 revision, this device is simplified by eliminating Section V altogether 
and coding general articles covering various processes and properties simply by 
notching A-Gi for “general metallurgy,” together with the appropriate Materials 
Index or Common Variables Index coding. 















































CLASSIFICATION FOR METALLURGICAL LITERATURE 


109 



Figure 5-3. Three examples of Materials Index Coding. Aluminum-copper-mag¬ 
nesium alloys indicated either by a combination of a deep hole in A1 and shallow in 
Cu and Mg, or by Al-h-40 (duralumin), utilizing the third and fourth orders section. 
Antimony, not considered a common element, is coded EG-a-31. 


Index (letters “a” through “s” and numbers 30 through 45). A completely 
coded outline for such subdivisions of the alloys of the common metals is 
provided in the classification outline. For example, the aluminum alloy 
known as duralumin can be coded Al-h-40, according to the classification 
outline. In Figure 5-3 this is notched deep in A1 in the Common Elements 
Section, and deep in h, and shallow in 40 in the third- and fourth-order 
section. 

This third and fourth order section is also utilized for indirect indexing 
of the less common elements which are not coded by chemical symbols. It 
will be noted that a double hole designated “EG” appears just below the 
center of the left edge. This designation (for “Elements Grouped”) includes 





110 


PUNCHED CARDS 


all the other elements of the periodic table, reached by combinations of the 
lower case letters and numerals in the third- and fourth-order section. For 
example, in Figure 5-3, the element “antimony” is coded by notching EG-a 
(deep) and 31 (shallow). 

The various types of steels and irons are provided for in the section at 
the bottom left in the card designated “Ferrous Groups.” The divisions are 
as follows: ST—Steels, CN—Carbon Steels, AY—Alloy Steels, SS—Stain¬ 
less Steels, TS—Tool Steels, Cl—Cast Iron and Cast Steels. This again 
constitutes a direct index. (See Figure 5-2, notched in Fe for iron and ST 
for steels generally.) 

For indexing materials by special properties or application, the lower 
left corner hole designated “SG” is provided, and is used in conjunction 
with the third order lower case letters. For example, heat-resisting alloys 
are coded by notching SG-h; bearing metals SG-c, etc. In these “Special 
Groups” fourth orders are open for further subdivision as required by the 
user. 

Common Variables Index. As stated above, the Common Variables 
Index makes full use of the unique faculty of a punched-card system to 
record several concepts simultaneously and thus provide selection of in¬ 
formation by an interrelation of ideas. It is a miscellaneous collection of 
factors which modify a number of separate headings in the other indexes. 

The Common Variables Index is accommodated on the right edge of the 
card and may be handled in various ways at the option of the user. Three 
methods, using either the direct or indirect coding principle, are suggested 
in the book containing the complete classification and explanation published 
by the American Society for Metals. 

Author Index. A new method of indexing authors was developed which 
represents an ideal combination of simplicity, ease of operation, selectivity, 
and economy of space. Studies were made of the frequency of occurrence of 
the various letters as the first, second, and third in metallurgists’ surnames, 
using the author lists from the ASM Review of Metal Literature and the 
British Metallurgical Abstracts. Results showed that the heaviest concen¬ 
tration of specific letters occurs in the second letter; in fact, the vowels 
A, E, I, O, U, plus the consonants H, L and R constitute about 90 per cent 
of the second letters. It was therefore concluded that, for greatest selec¬ 
tivity, it is even more important to index the third letter of the author’s 
name than the second. 

The Author Index is accommodated on the bottom right edge of the card 
(see Figure 5-4). The first letter of an author’s surname is indexed in the 
deep holes in the larger field designated “First and Third Letter.” The third 
letter is indexed in the shallow holes in this field. The second letter is in¬ 
dexed by a deep or shallow notch as required on the small field designated 



CLASSIFICATION FOR METALLURGICAL LITERATURE 


111 


(mI) 


* 

n 
a 


I 


1 

IBaHMMtMIIIMMU 

n ft? 

1 


1ST AND 3RD LETTER 

210 LETTER 

zq 


AUTHOR INDEX ] 





Figure 5-4. Indexing of two authors on one card. Wilson, coded W, I, L, and 
Adams, coded A, D and DUP, for duplication of the first and third letter. 


“Second Letter.” In this second-letter field the vowels are each indexed 
separately, while all of the consonants are indexed by notching the shallow 
hole marked “Other.” 

The common combination “Sch” is considered as a unit and indexed 
separately. The following fourth letter of the name is then indexed in 
the second-letter field, and the fifth letter is indexed in the proper shallow 
hole in the first-and-third letter field. 

The shallow hole marked “Dup” (corresponding to the deep hole “Sch”) 
is provided for instances when the first and third letter are the same. 

Capacity for Expansion. It will be noted that there are two shaded 
areas on the card (Figure 5-1), one in the Processes and Properties Index 
covering the code letters W through EE, and the second in the Materials 
Index, letters FF through NN. These are provided for addition of any 
subject as required by the future growth of science and technology. They 
may be used for the addition of material in fields other than metallurgy, 
or for advancing specialized fields to more important first-order positions. 
For example, a metallurgist who has a large file of references on electrolytic 
refining may wish to advance C-23 to create W — Electrolytic Refining, or 
one who is particularly interested in materials handling may save needling 
time and provide for more detailed subject breakdown by assigning first- 
order division X to materials handling instead of notching and needling 
through A-5-a. 

The potential capacity of the classification system is indicated by the 
following figures for the Processes and Properties Index alone: 29 first- 
order divisions; 27 subdivisions under each of these, making a total of 756 
second orders; 16 subdivisions under each of these, or a total of 12,096 third 
orders; and 15 subdivisions under each of these, or a total of 181,440 fourth 
orders. 

Some of the open areas will be used by the Committee now engaged in 
revising the classification. For example, in the five years intervening since 
the first publication, experience has shown that certain nonmetallic ma- 




112 


PUNCHED CARDS 


terials used in metallurgical processing should have some place in the out¬ 
line, and a new section in the Materials Index will be opened for this cate¬ 
gory. Many new subsections will also be opened under the various main 
divisions of the Processes and Properties Index and the Common Variables 
Index. A good example of the capacity for expansion of the system is the 
fact that all of the metallurgical aspects of nuclear sciences (w'hich were 
unforeseen five years ago but now constitute a vast and widely assorted 
segment of the literature) will be accommodated in the revision merely by 
adding appropriate subdivisions to the various sections of all three indexes. 

The Workbook. Since one of the most useful features of the ASM-SLA 
classification system is its capacity for expansion, a highly adaptable method 
of working with the classification outline is demanded. To make most effi¬ 
cient use of the classification, therefore, a looseleaf “workbook” is provided. 
In this book each of the main divisions of the classification is printed on 
looseleaf sheets that can be thumb-indexed for ready reference. Existing 
second- and third-order subdivisions are also printed, together with a com¬ 
plete list of coding symbols for all of the open second, third and fourth 
orders, thus providing space for additions to the classification at the will of 
the user. A sample page from this book is reproduced as Figure 5-5. 

Experience in a Company Library 

In early 1950, a new Research Library was being organized at Kaiser 
Aluminum and Chemical Corporation’s Division of Metallurgical Research 
in Spokane. At the same time the American Society for Metals and the 
Special Libraries Association were jointly preparing the punched-card 
system for coding metallurgical literature described in the first part of this 
Chapter. The need for an abstracting system in the Kaiser Research Li¬ 
brary, coupled with the Librarian’s previous experience with punched cards, 
made this ASM-SLA system seem desirable. Furthermore, this system could 
be readily adapted to a young organization such as the Kaiser Research 
group. These factors guided the decision to install the system as soon as it 
was completed . 8,9 

Word reached Spokane that the ASM-SLA Classification would be 
available at the time of the Special Libraries Association Convention in 
Atlantic City in June, 1950. Of the two copies of this Classification of 
Metallurgical Literature which were available there, one went to the North¬ 
west for study and adaptation in the Kaiser Laboratories. 

In mid-July, 1950, the various items needed to start the punched card 
system began to arrive, and abstracting was begun on all literature in the 
Library, except the books and periodicals. In about two months there was 
a fairly large file of technical information recorded on classification cards. 
Then the work of actual coding began and many hours were spent going 
over the cards, tentatively coding them, and then exchanging ideas between 



Third-Order Divisions 


B — Raw Materials and Ore Preparation 


10. Ore deposits & mat. res. 

a. 

b. 

c. 

d. 

e. 

f. 

g- 

h. 

j- 

k. 

m. 


P- 

q- 

r. 

s. 

11. Sampling 

g- 

h. 

j- 

k. 

m. 

n. 

P. 

q- 

r. 

s. 

a. 

b. 


13. Crushing, grinding, and sizing 

a. Primary crushing 

b. Secondary crushing and grinding 

c. Milling 

d. Screening 

e. Wet sizing 

f. Air classification 

g- 

h. 

j- 

k. 

m. 

n. 

P- 

q- 

r. 


14. Concentration & beneficiation 

g. Gravity 

h. Flotation 

j. Magnetic 

k. Chemical 

m. Agglomeration tabling 

n. Electrostatic 

p. Settling, thickening, filtration 

q- 

r. 

8 . 

a. 

b. 


c. 


c. 


d. 


d. 


e. 

f. 

12. Mining 


e. 

f. 


15. Roasting and calcining 


n. 

Open pit 

n. 

P- 

Underground 

P- 

q* 

Hydraulicking 

q- 

r. 


r. 

s. 


s. 

a. 


a. 

b. 


b. 

c. 


c. 

d. 


d. 

e. 


e. 

f. 


f. 

g- 


g- 

h. 


h. 

j- 


j- 

k. 


k. 

m. 


m. 


Figure 5-5. A sample worksheet showing blanks for additional entries in third- 
order rank in Section B on Raw Materials and Ore Preparation. 


113 



114 


PUNCHED CARDS 


library and technical staff members. There were many conflicting ideas on 
how to code, what to code, etc., but after checking about a thousand ab¬ 
stract cards, things began to smooth out and the system began to take 
shape in the Library. 

In many organizations, where an abstract file is already in existence 
transferring the information to the new cards would be almost an impossi¬ 
bility due to the amount of clerical work involved. Microfilm strips con¬ 
taining the abstract could be inserted in the cards as a timesaver, but would 
necessitate special equipment, such as a reader, special cards, etc. In a newly 
formed organization, starting from scratch, the installation is somewhat 
simpler. Consequently, it was relatively easy in the Kaiser Laboratories. 
Many industries might find the ASM-SLA Classification easier to use than 
do those in the light metals field. In dealing with the aluminum literature 
it was found necessary to do a great deal of expanding into the worksheets 
provided with the ASM-SLA Classification books. 

The ASM-SLA system was adopted for indexing all reports on research 
work conducted in the Division, as well as supplemental information con¬ 
tained in published articles related to similar projects. Thus the punched 
cards in the Kaiser Library now include references to papers from many 
sources such as the Bureau of Mines, U. S. Bureau of Standards, NACA 
Reports, PB Reports, etc. 

No attempt has been made to code books and periodicals, because they 
are cataloged according to the Library of Congress classification. Of course, 
some technical articles in the periodicals are abstracted and incorporated 
in the ASM-SLA system. 

As a safeguard against overloading the cards with extraneous material of 
little value, Kaiser research staff members help the Librarian to select 
published articles and abstracts for inclusion in the file. A number of the 
ASM-SLA Classification books have been assigned to department heads for 
use in coding reports and articles. The ASM-SLA “Worksheets” are also 
provided for each department head. 

As new subjects are included in the system, the items are added to the 
list on the Worksheets. A supplementary index is issued when enough items 
have been submitted to justify the publication of a new list. Copies are is¬ 
sued to all members on the staff who use the ASM-SLA Metallurgical Clas¬ 
sification. 

Figure 5-6 shows a page in an abstract bulletin with code annotations in 
the margin and a card punched and abstracted accordingly. Only the 
“Processes and Properties” and “Author” indexes were needed for this 
particular reference. Of the code symbols jotted in the margin, Q24 at right 
stands for Plastic Deformation (24) in the Mechanical Properties section 
(Q). Q3 stands for Creep, M26s for Crystal Imperfections, and M22g for 
X-Ray Diffraction. 



CLASSIFICATION FOR METALLURGICAL LITERATURE 


115 


THE KSCHANlSn 07 CfiJOP AS RX7ZALS0 BT 1-RAT «TH 0 D3, 0. B. 
Orwwnougb end 1. u. Swltb, Journal of tbe InatltuU of rtetmle (Xnslend) 
77, part 3, pages 433-4 . J, July 1930. A hypothaaia la put forward to 
•xplaln In terse of dlalooatlon theory the recebt obeerretlone of Vllse 
and Wood (J. Inat. Hotel*, 75, 693, 1946-49) and of Wood and Raohlncar 
(J. Inat. Hetele, 76, £37, 1949-30) In ralatlon to tha aaohanlaa of tb# 
deforsetioa of watala. Tbla aachanla la eonaldarad to ba oloaaly allied 
to that of tba polygonlzatlon phenomenon obeerred by Cabn (J. Inat. 

Hatala, 76, 121, 1949-30), and tba bypotbeala Innnrpnrgtm r~~-" L ~Tl»r 

ldaaa. Some naw X-ray .. I I iiiiiiaMi _ ~~¥ 4, a a'- 

tal evidence in famia^f afee^iz>? * • • • • i * 4 v ^ ^ * a «•!* 




• * * 


m ^ RiVEAlfl* BT ' ErHCM 

“ KlUHI3i 35 ^, 3 . (July. W50) 

ws* sr-: "• --rssot 
«*— ta r,v 







Figure 5-6. Page from an abstract bulletin with code notations in margin and 
punched card slotted accordingly. 


Occasionally, an investigator finds that the standard classification does 
not adequately index a particular article or report. When this occurs, he 
discusses the matter with the head of his department, and the two of them 
select a new subheading to meet their needs, decide under which of the 
main classification headings it belongs, and submit their suggestions for 
expansion of the index to the Librarian for final approval. The new sub¬ 
heading is then entered in all copies of the Worksheets. 

For example, several new classifications were needed in connection with 
remelt studies. In the aluminum industry, alloying and ingot casting are 
not considered part of the refining process as they are in some other non- 
ferrous industries. The nature of remelt practices indicated that they could 
be added best to the section of the “Processes and Properties” index desig¬ 
nated “E—Foundry.” Three new subheadings were entered in the Work¬ 
sheets—namely, “E7. Ingot Casting,” “E8. Fluxes and Fluxing,” and 
“E9. Alloying.” 

Because of the uncertainty of fields and phases of research in which the 
Division will be engaged ten years hence, it was deemed advisable to mini¬ 
mize the use of fourth-order indexing and new headings. Fourth-order 


116 


PUNCHED CARDS 


divisions are added only when unavoidable; consequently there is still 
much room for expanding the system. 

The “Materials Index” section of the classification provides a good break¬ 
down for coding aluminum alloys. With the addition of alloy 50S (K-150) 
as “Al-g-34,” this section is adequate to cover most existing commercial 
alloys. The “Common Elements Index,” however, is used to classify ex¬ 
perimental alloys according to their chemical compositions, as well as for 
other metals and alloys that have no commercial designation provided in 
the standard index. 

The “Common Variables Index” may be used to code information on the 
construction and use of specialized laboratory equipment, or to separate 
reports on different phases of a research program according to the influence 
of various factors. Although the entire section will be used eventually, the 
divisions which are currently finding most use are those numbered 1, 2, 3, 
4, 10, 11 and 12 (equipment and processes, influencing factors, wrought 
metal forms, type of literature, form of literature, and language). 

If a literature search is made by a staff member, he is asked to furnish 
the Library with a copy of every abstract he prepares. If it happens to be a 
very extensive search, he is requested to compile a bibliography and an 
MR (Miscellaneous Report) number is assigned to this particular search. 
This bibliography is then coded for the subject matter and is also coded 
11-15 to indicate “bibliography” in the Common Variables Index. 

From a time-saving viewpoint, it is necessaiy to keep the cards in a 
readily accessible filing cabinet. Since no filing is required, this system can 
be used by technical men in their personal files. Of course, the coding and 
notching of the cards must be kept up-to-date or the system defeats itself. 
Furthermore, the cost of the cards themselves is too great for them to be 
used merely as regular abstract cards. 

One trend which can cripple the entire system is the inclusion of too 
much material of marginal interest. If the people on the technical staff 
proceed to code all literature they may encounter, the Library will find 
itself swamped with journals of no probable lasting value waiting to be ab¬ 
stracted. However, an explanation of the requirements for a good abstract 
file will generally result in full cooperation of all participants. In order to 
build a good technical library, the wholehearted support of the staff is 
necessary. As Ralph H. Phelps of the Engineering Societies put it during a 
New York SLA Meeting, “know what you need and get it if available” 
with the corollary, “avoid getting and keeping what you do not need.” 

Another question that arises in connection with the system looks to the 
time when the number of cards will be too cumbersome to handle. There are 
several solutions to this problem as we see it now. One would be to weed out 
abstracts covered by the various abstracting services, such as Chemical 



CLASSIFICATION FOR METALLURGICAL LITERATURE 


117 


Abstracts, Metallurgical Abstracts, etc., when the annual indices are re¬ 
ceived in the Library. Another practice would be to subdivide cards by 
subjects as listed in the ASM-SLA Classification of Metallurgical Literature 
under the “Processes and Properties Index.” This, of course, would mean 
needling through about 20 to 25 sets of grouped cards, but it might be 
time-saving in the long run. An efficient mechanical sorting device might 
well be the ultimate solution. 



ASM-SLA CLASSIFICATION, INTERNATIONAL (SECOND EDITION), 1958 


118 


PUNCHED CARDS 


ID 

a 

o 

I 

W 


X 

0> 

3 

O' 


bfi 

c 


63 


13 

a § 
^3 bfi S 
o a 


bfi 

c 

*3 

X 

Cj 

03 


£ 

.2 s 


52 % 

o3 a) 

O > 

bfi 03 C ~ - 

.5 ° * s *- 2 

13 g ° ^ *g 

rr5 c x 4> c 03 

ij £ 43 -j? 03 2 ~ 

Saj&,aoa-o 


.2 5 


X 

13 

O 

-C 

0) 

s 

x 

o 

J3 


bfi 

G 


c 

O* O 
Si « 


bfi 

O 

o 

c 

« g> M> <a> 3 

-x> .g g C . c C =3 

T3 3 j) *Sq C V2 ““ 

2 "o S .5® § § ® 
CQ S o cu o S 


§ 2 

■S'6 


Si 

o 

•<* 

*•<» 

o 

S 

© 

"»«* 

x 

x 

V. 

o> 

si 


o3 

J2 *5* 


oo 

O 

o 


SSSJSSSSSSSSSSSScSSSS 


bfi 

C 


o 

!S 

53 

’£ 

c3 

,G 

o 

43 

S 


63 

s 


bfi 

G 

£ 

a 

IS 

X 

o 

TJ 

c 

63 


X 

u 

o 

43 

£ 


6 

63 

-C 


X 

•© 


13 

si 

C3 

oo 

V 

A 

CJ 

-«J 

oo 

if 

© 


s: 

o 

5 £ 

x a. 

H -§ 
43 ac 

a. 

43 s: 

H ^ 
43 13 

c 


*ts tc 


e3 bfi 

S.S 

a be 
« 2 
a 


c 

o 

SP*S5 
£ 3 
53 £ 
O X 

55 W 


.5 c 


be 


^ - = 
.5 S •= 

?•! « 
mH2 


S S §5 c5 S3 c§ Sn 


43 

a 


3 

o 

~o 

a> 

oo 

a 

a> 

13 

a 

o 

Q 


.s 


x 

■o 


8 

€ 

• 

a 

2 

a 

"2 

o 

8 

8 




bfi 

.£ 

‘5 

<e 

0) 

55 

TJ 

G 

03 

fl 

.2 

x 

63 


X 

W 

i 

o 



bfi 

C 

15 

c3 

s 

'S 

43 

-fcj 

vi 

13 

3 

63 

a 

o 


2S8S8SS5SS 


»c <o 


o 

‘3d 

X 

^3 

a; 

s 


4) 

C 

a? 

a 


S3 

O 

£ o3 
2 § 
•2 -o 
EC W 


S 

2 ao 

g ^ c 
g si o 
8 * - 
13 

a 

63 


§ « 


2 O .5 
•5 ft £ 

OO X 

w c _2 

03 63 “O 

^ r c 
!/2 pu, HH 


o 

a 

'CC 

"S 

4» 

•»— 

00 

CJ 

s 

13 

Si 

C3 

. Si 
>» O 
•»* 
03 

<41 2 

o3 

® "o 

13 A 
G "S 
.2 

-A £ 
<*-> 00 

Oj "S 

£ c 

32 s 


e oo 
o 


I 


O O 
oo 
o> 

t. 


& 

.A 

1 Si 
C3 .O 

-G *3 

P s 

c3 ^ 
03 5 

x s 

03 o 

P5 O 


x o» 

05 Si 2 

•S -2 8 .? 

^ m ^ -C 
£ ° '2 c 

W — g 03 

| il-s 

•h?l 

g« 

,03 


oo g i 
o' s £ "2 

^ O •£ 5J 

Oh *3 *4 ^ 


CsjcO^^OSOI^OOGa^C^CO^fiO 


o 

CJ 


PQ 


■t- X 

M 50 

CJ 03 
k S3 

a o 


8 5. 
V c 

O O 
£ - 
ft s 

o ~a 


S3 03 

£ £ 

£ * 

92 X 

a 03 
O PQ 


.2 .2 S ^ 

U I-. “* 

C3 03 
03 03 

s s 


X 

03 

X 

X QQ 

03 S 
g X 
P 00 

a £ 
Sv ^ 

3 

-2 -2 
2S S3 

*2 8 . 

S CQ 


O Oi 
»-» v* 

^ X 
03 

5« S 2 

s c « 

*5 o ^ 

S g g .5 

^ 03 

o ^ «0 a 

M ^ 5 

c H ^ 

ioa. 


^WW»0<ON00050-n 


G 

O 


63 

a 

43 


U 

O 

"S 

s 

o3 

55 

ID 

G 

63 

43 

k* 

o 


PQ 


bfi ^ 

c s 


» 03 

*2 ’o 

g «a 

63 03 

c ® 
m .£ ^5 

-S T3 w 

O C G 

js *C o 

bfi 'Z 
£ bfi t - 
bfi.£ G 
G -G ® 
■G “ 2 
o G G 
u O 

o o 


.o 

o 


o 

Cft 


w 

4-> 

43 


X5 


bfi 

C 

‘H 13 
*S 00 G 
*2^0 

S 00 si 
u £ § 
TJ § vs 

cj a ^ 

bfi Oj £ 
G .8 .2 
w ^ 13 
2 £ — 
o * § 
55 d3 ^ 


43 

*n 

o 

8 a 

> C3 


43 
G 4) 


S3 

43 


1? 

si 


© © c? 5J ® 

I? §“1 ? 


xxx 
xxx 
xxx 
xxx 
g ® O 


a. a, a, ss 


(NW'tiftCONOOOJM 
i—(i—lr-(^HrH^Hr— 


*c 
CM CM 








CLASSIFICATION FOR METALLURGICAL LITERATURE 


te 

a 

cc 

'o 

a 

C bC 
03 O 

e * 

-5 o 
a o 

®3 

^ I 

be 

•a fe .s 

O to ■*-> 


S 8 

o ° 

ii 

f j •* 


bC 

S 

’■w 

ce 

o 


Os 


be 

c.S 
§ -2 « 
I'S S 
* a.s 


a 


o be •_ 
« a " 

G 

2 S o 

§ 8 *- 


* «, 
.5 .s 


j] 


■S"« 


S 2 
” "a 


J J *5 & ® 

H O Q Q S 


e § 

£ s 

2.2 

« § 

3 ^ 


■■s * 


Ml 
C 

w w 09* 

.« Ml Q, 

*§ .s » 

J 3 « 

•5 12 “ 

oj J? 4> 

oSS 


be 

a 


* -2 
s o 
o-§ 
£ £ 

*> o 


a a o 

g ?TS 
3 Ss 5 
Os Os aS 

^ be 

Os Os C 

•r* ■<* 

^ c 

o o 

O © aj 

O O Pu, 


be 

c 


•5 -3 


el 

g | 


*> 

5 


S 8 « 
o o o 


W^*»0®NQ005'HNw^»fitOSor)0> 

^ ^ »—• *—«,-* i -^c v ic s jc'$c ; ^c s Scnc'i65cni 


G 

o 


•3 8 

S 5 

o « 

o 2 
£3o 
■f fr 

M>| 

a *c 
■3 ft. 

® e 

s § 


k. 


m 

£ 

3 

*-> 

a 

3 

k. 


2 

a 

£ 

A. 


00 

no 

o 

M 


O ZL GO 

S S i 5 

&§.S ft’s 

8 -5 ® *>5 ° 

e oo o +* ^ o 

.5 O rt ^ 

V O Its A 2 § 

A. JZ •s ±> j3 ■** 

ooSQO^t 


2 


00 

£ 

3 

o 

£l 

3 g 

t; 2 

3 .g 

-g i 

ao .. 

_* Os 

13 h s 

^2 O O 

El | 

a ft, S 


£ 

3 

a 

2 




be 

a 


a> 

AS 

•3 

a 

o3 ao 

« £ 

g 3 

O -k> 

2 2 

i s 


a 

as 

Li 

H 

I 


s 

® 

e 

I 

H 

3 

O 

s 



00 

4> 


a> 

2 

a 


<n 

"3 

+-> 

V 

B 

a 

o 

a 

o 


Os 

£ 


93 a 
*** *-2 .3 


be 


O 08 
k. 

*3 0 ) 

I e 


4? 13 bC be 

& 2 fl o 

® 3 *c a S o 

•3 i j&jjj •§ |1 

pSHf2oc8fflif§ 


be 


3 00 

s s 


^-3 
g .8 
> G 
00 2 
4) p3 

^ 8 




HN«^»O«N00OiHNCO 


be 

G 

IS 

ao 

;s 

£ 

*3 

a 

08 

be 

a 

*w 

as 

O 

O 

be 

o 

3 

08 

a> 

O 

I 


be 

3 

IS 

k. 

O 

£ 



> -S ‘q. .£ C O T 3 .5 a 

« h c £ ae^S^-S*^ S a *a 

S 3 § 5 a g g 5 =s £ a J? 
£5 5 A g ©i® Oja’S,! 
acQcoCsSpQojMfSHccPQ 

•h ci po uj ^ oi o th csj w 



.5 ¥ q c .5 

*3 


te 


08 

I 

I 

se 

o 

Ah 


NH 

CS HH 


OHMM^«S»05NWi' 


be 

.5 

*<♦3 
be ^ 

a S 
3 ao a 

,2 #TJ fl 

g I 

*V Su-° -w 

2 s a § 

a a | © « 

u u ° W) G 

^ be c . c 

•g -g .2 5 5 
> > x *S -S 
O Ogi.S 

ft. ft. S S CO 

o fH ci hj< «j 


119 


16. Post-sintering operations 10. Mechanical cleaning and polishing 1. Diffusion 

17. Fiber metallurgy 12. Chemical cleaning and polishing 2. Nucleation 









ASM-SLA CLASSIFICATION, INTERNATIONAL (SECOND EDITION), 1958 


118 


PUNCHED CARDS 


o o 

a- 2 

1 M I 


O 08 0) 

to S u > 

® .3 - »*a ® 

SC g 5 9 *= .2 

.5 2 g ‘C » 

^ -o S 5 

.® £ «5 S £ 


S tf « M m a, 5 

+* .5 g fl C £ 3 

-o 2 1 ’Sb *5 5 

a *o h 2 2 o 
« ig © a P P *S 


Swa,Qoa,Ofi.i»So«a.oSo 

ssasssssiassaaaas 


hols gf 
.3 S .3 

b© £ 2 
c3 -Q sS 
j; 3 >- 

02 H G 




! 2 c 
» £ *> 

> o 5 
! ® 5 » 

! 05 Q as 

+* 

| §■ | 

ir-ffl 

>* - *tr o 

^ 5 ° •- • 

. o _ . 

q 'c g « • 

! .o o .2 6 

u ftg S 
‘ ~ S g "oj 

a; £ S - S 

<J < 


0) C 00 

§.£ S » 

o, a ? » 

B 

,22® 
«-S & S. 
.2,2 5. 

^ ,«o O 

*• 

o o ^ ^ 
4) © 3 § 

S 5 ^ so 


Mg I 
A 8 < 

V3 v- w 

S - s .£ 

o ^ « e 

g H 5 S 

H O ^ 


2SS*S3cS8c5c3?i 


NC0»0©N»0>O^ 


I *8 

n | 
. 2 I 
§ >>.® 1 
~ £ "3 -o 

W gj ^ 

•T’ll 

- 30 ^ 

.5 o3 .« -G 
erg 

*» J3 C 2 

3D 4-> flQ 05 

m I 

►5 sc * « 


■ I « 2 
^3 r 


to § 

B C5» 

1.1 
•31 r 




5 '•5 ^ S « a a « i: 
ocq^g.©-*-* 
si 'C o -o 

g 

S b© £ be £> -C .« .C .S 

tu .5 b .s -5 «5 55 55 

|-S 2 ss’ 8 2 S 

•S S § o * S S S S 
SooiatoCa,a.a, 

{Noo^idcdi^ooaioo 


sj 

32 03 

«±5 

5 £ 8 "g 

® v 5 

v^5 C5: Cs> 

c c c 2 

**■«» •*- •*« v 

30 «C «C ^ 
00 30 cc , 

V V ^ 

O ^ 

8 2 £ « 

ft, a. a, at 








CLASSIFICATION FOR METALLURGICAL LITERATURE 


119 


§ g q 

S* 

| • s « 

5 m q S sis' 0 , 
o o bfi *g - s fc 

"3 c ‘2 * 5,3 8 .£ » 

« .2 2 p g « •« .5 

?«.&«« § *«, 
OQQWHtOO, 


o o o 


2 « « 
o I « 2 

ea +* 2i 2 "Si 

i i.iii 

L . r} 

»• £* a xt je2 

.* r O o < 5 
55 © •+* *_ft © 

S « « 

,5 o oJ ^ $ 2 
C K J* ® 2 6 
v © *2 33 2 S 

«sS <5 o m -c 






OQ 

q 

R 

q .2 

aS 

2 ti 

H 

S| 

1 

to § 

55 

5 ^ 




© 

*© 

h 

flS 

5 **> 

q 

2 ^ fl *3 

b© c P 

ft .3 0) m 

b©3 bfi §> ’g Tg ft 
ft .2 ft 55 _S i •? ‘ 
q .3 ‘i 3 a U 

i "3 2 a "v S'- 0 £L‘ 


S'g'g 1 ai:s-1 1-2-1 w il 
sass§ss?as' , '" e,i ^“ i 


9 m g 

n 3 ft^ 

|2 9 B 

gsJlIg 

'-3 '-3 e 1 £ ® 

2 - § % Ss * 

> ^ .2 § & « 

O CD 00 © u A 

u « © A O ^ 

<5 O PS H fa O 


© 

s 

S 534 

2 * 

q ^ 
o 

^ bo I 

g 5 ;§ § 

g g*2. « 

fe£ § M o> w 

fl » P ft 

t fl 8 .2 ’S § 

© *2 $ *© » *q 

2 S S a S 8 

■s c S o-si 

a§ pa t£ pq ■< S 


*3 w 
M ft 
.2 *3 bO 

ft *g ft 

t£ ;s .o 
^ o ® 
-V CU^ 

§ *§ 8. 
m § 2 
•I *?§ 
8 -s? 

O rt *3 

a> c 
- ®3 

be w «> 

.2 ”3 ^ 

§ -2 *3 

2 S-l 

O -a S 

I § M 

' So 




„ 1 
o> ft bfi ° 

st .5 q ^ 
*? Q,*2 « 

& a 8 § 


C w 0© » h r* 

.Sq-MQ^aqo'O.S 

a -r « .2 o ft os o 

lU^£«=£i 


bC 

I si 

a '& ® 1 


g Ji « C a « £ « o j: '5, 3 'i « “ L- r r." „?■ 
£^35c5mMDBtfH » a QSot.mfc.tc 

-!ei«^ui®o>o^2jS2S2tiS22?j?jS 


be 

a a 

•2 .£ 2 
■P J3 O 


o r © 

T 3 Se" 0 *« 

2 211 

cl o. h te 

“ r ° be * 

v S Sf.2-C 

-a -a .2 -u v 

% | .2 -5 a 

p. p. S S in 


o - « <ji « e n 









. Grain growth 28. Compression test 21. Miscellaneous service testing (life test- 

. Recovery 29. Hardness test inir) 


120 


PUNCHED CARDS 


o o 

X 







TJ 

a 

2 





si 


S3 

*3 





V s (L. 

c £ 


o 

• n» 

o 

u 

a 





*f»» w 

00 ki 

2 'v 

"3 

*«*» 

k. 

o 

Q 






Cn ^ 
© § 

^N» 

k. 






v- K 

3 

ft £ 

-4-a 

0> 

> 



X 

ft a 

S £ 

TJ -2 

In -** 

s 

VS 

Q 


"2 « 

£ 

•*** 
c3 e 

» s 

O 

08 O 
■gso 

1 

S 

o 

o 

c 

•r- 

o3 uh 
O 08 

In, In, 

3 

a 

a .<-> 

. N ’ W 

6 * 

Pi *3 

S 2 


0 g -o 

I fc-l • p* 

! 03 

+2 o 
m O 


= JS * 


OS • "I 

ft ^ 

c si 6 

oo IS ° g 

c x 3 

S « - S 


<Stf<!ajOE*5C)<O^S*<U^ ChQOS 


g .2 >, 

a> a 5f 

fi c4 O 

* & g 
S o 


«r ^n» it 

OO Aft ®Q -*n> 

*>.**■*» g V « 00 

^ * £ oo * ^ M 'S 'SS 

| s s-i |.l|s t 

® S 

HOTQPQ®taa^O 

^Nco«5«dNoooid 


i « 

CO •« 


g 3> 
Sag 
- ^ .2 
5 c -p 

O o 

“■5 3 © 
, *3 .2 o 

5 O ^ *U 

a .2 ^ 

c ^ c£ 

g cj d 
O GO P* 


h^P5^®N«0 


I § S 

0 J o 

§ * •£ 
H 0 o 
C ^ o3 
O c3 4> 
«—. •+-* *-. 
x 

O • T3 
.2 "C £ 

"5 O 
O x 
o3 x 

»- 12 

S..2 g. 

to O J 


c c c 

!3 O ,H 
fll .- t* 


u* -*-» 

^ u 

W 

c3 

35 



E 

fe "1 

ft g 
C c 
v. 

c3 

~ 4? 

& ^ 

0) 

In, 

•a 

JS 

a 

£> 


ci 

C 

^ J3 
35 O 

•^1 X 

’3 

£ 

i 

ik 

c 

3 

"3 

a 

J W 

o 

a 

ft a 

.2 
4> In, 

H 


rf oo oo 
.2 3 JS 

ij e> cs 


^ o t: 

,n_4 *.£ 0> 

c3 u a 

o o o 

‘3 ft 9 

5 o a 


CO^ICCO^OO 05 O h N ifl (O 


® S ©t? Ml S ^ 

* J= NC 3 i * « s 

ShHccW^cS< 


W CC lO CD N OO 


23. Plastic properties 14. Size, thickness and mass measurement 6. Mineralogy and petrography 

24. Plastic deformation mechanism 15. Surface roughness 7. Engineering 

25. Stresses 16. Temperature measurement and control 10. Applied mechanics 

26. Fracture mechanism 18. Process control and measurement 11. Stress analysis and elasticity 

27. Tension test 19. Radiation detection and measurement 12. Biology 











CLASSIFICATION FOR METALLURGICAL LITERATURE 


121 


to 

■e 



£ o 

^ to to ftq ^ a, 03 

—I oi w ^ to 

M M r -1 *-< i-< , ,-h f-i 


a 

• m 

*i 

c 

cc 


S ^ S a § 

3 § .3 © *2 

*«* 55 ;** ft 

^ •-* ^3 QQ ft 

3 S § « o g 

."S a e 5 c to 

to q 3 *5 > 

n 

^ ^ g£ CS3 N 


5s 

~ .3 

§ s 

1.1 

*♦* 2~ 

3 00 
3< -*s 

V 2 * 

3> Q & 

.5 -5 

45 3 * 

w S n S 


a 

3s 

■S ° 
-5 

•2 © 
3 -S 


5ft 

•S 00 

2 2 
to 3 
© © 
£ t: 

ft. CL 


&! 
© 52 

to « 

S* as 


- -O 3 
Q * .« 
« c» 2 

fe |J 

to 

•<£ to to 

©S*^ 


#>* V 

PI ^ 

3 3 

.g 2 

3 to 

3 

■S "« 

to 3 
^ 5 


5ft 

ft 

-ft 

to 

•« 

ft 


3 2 


2 

K> 

3 _ 

II 

is 


.*£ 
"2 ft * 

5 S 

'totS 

"2 *• 

3 to 

© < 


3> 

to 

3 


*0 ft 

3 to 

^3 ft 

i - 

i g. 

o &q 


■§ 1 o "5 


2 _ 

§ •§ 
5 ,® 

3 ^ 
£ "3 
2 § 




3 

s 

’■s 

ft 

3 


to 

3 


"to 

8 S’ 

3 P 


oq m 

.2 2 
■S •** 
k y 

§. s 
s §■ 

«■ ft. 


■< 2 

- ft, 

to 5 
M ft 

<5 o 


a. , 

3 .© 

to ^ 


2 2 
•&! 
$ s 


. to 

■s I. 

to 3 


to. 

8 . 

a 

o 

O 

I 

3 

u 


i 


s 

a 
a 

<D 

*0 

SS-o § 

s s 11 j i 

i i i i H "" 

be a 

s s 


s 

.3 

*53 

o 

a 

$p 


S 

92 

0> 

ft 

ft 

bO 

ft 

ft 


Hit 

g £ £ •-- 


c n 


O N OC CD H N W 

3* cs c$ 5l 


S 53 85 SI cJ ^ 



3s 

to 

ft 3 

•2 ft 

«= I 

3 to ft 

’^2 © to 

k* Q_ *r* —» 

ft to '■-* 3 

ft 

to 

5 3 

^ E to 

• 2 - 

■2 * e | 1 « 

3i 

kq 

^ "§ I -2 S. | 

c 13 .§• 2 •§ -~ 

to 

ft ft 

5 .2 

« ?g, 8 . § | 

^ © 

- i ^ g «■ -S TS 
3 S ps -* ft 


3 » 

-ft u. 


3 to - 

>* g ^ » 

2 to ^2 ^ *3 

•5 ©» .© -3 ft 

? 1 c 5 55 

5 I 

ft. ft. 


v. ^ 

to 5 ^ 

3 2 3 

^ to 3 

^ O O 


oo 

to to 
12 to 
T3 3 
— ft 

to t* 

3 3 


3 

O 

to 

*■3 

ft 

3 


3 

ft 


3s 

ft 

ft 

<«> 

to 

ik 

3> 

©» 3 
to 5 
3 3 


ft 

3 


3 

ft. 

to 


-O 

3 


3s 

,ft 

ft. 

3 

-ft 

to 

>3 - 
ft © 
3 S 


*3 3s 

2 .g ft .g 

to to 

S | £ § 


^ to 
^ 3s ft ft 


T3 
ft 

° * ft - 

1 ss e -g 

? 9 O ® 8 

ft fc. as =«. ^ 


92 

>s 

O 


o 

p£3 

H 

T3 

C 

3 


I L T i 


CL> 

B 

V 

s 

a 

o 

S 

S 

o 

O 


| s 

c -2 c 

fi "ft, o 
3 K ■£ 


oj * 

> 


to . 

< < QQ O 


I a 

•2 3 

1 3 a 
I i 2 

3 © -C 

m 

o o o 


6 h w w «5 o n oo* d 6 n c 6 ^ ui 

*-• —i ^ ^ h 55 6i w w <n 






122 


PUNCHED CARDS 


OP oo 

* S 

Si Qa oo 


3 II 


•I ® "I 


si 

V 

s 

| si 
■ il* 

2 Qj 
5 o o 

| £-8 

I !j 

p£3 ► 

S | ^ 

S J ^ 


O O 

5 69 1 


w. <j 

O WJ fe. 


o « w w "r ifl e 


U)M S fi d d* 


C & ? k (a e] 


2 r ? 

°-§. I * 

J 1 a --g-s 

^ *2 «p g 

v ^ § 1 ® 

« & « b ■§ 5 

i ^ 2 § S I. 

n § | a I § 

O * s J! -3 

® tsli 

sc *2 tt; *2 i 


•2 

.2 -2 a S 


G D. C u xt c5 


'8 .« 5 S -i -8 8L S 

§* 1 ~ I 1 I * -c 

c5 JD 6 T3 6 s~* bio -C 


v> V 

•tt o » 
« ^ e 

o &5 *5 




— -O ® 

*f If 

O w gl 
w S 


G 

o < 


m 

o ^ as 


© 

4 C 

11 

■5 * 

* I 

S’jS 04 

all 

3 v 4 

8 s 5 


•a as 

£ 03 


^ C |3 

18-1 

© o,T3 o 

b’3 2 °* § 

•S cr - s § 

£ # * » 3 

?s£ S « 
V ® *, -S -s 


IL. 

* 


3 v <5 5 s «s «5 

> 4 8 SL § - afc 

g^HoQOfiQO 

Q. 

s W O h N « 

i* - * fH ^ 


14. Acid 24. Ultrasonic 17. Effect of trraaio 

15. Basic 2. Influencing factors 3. Influencing factors 

16. Hoi 9. Effect of grain size 16. Effect of stress 






17. Effect of time or rate 19. Cemented carbides 23. Overheating 

18. Effect of deformation or strain 20. Metal ceramics 24. Distortion 

19. Effect of impurities 21. Porous metal parts 1. Impurities 

20. Effect of prior history 22. Structural parts 2. Scale 

21. Effect of prior structure 23. Briquettes 10. Type of literature 


CLASSIFICATION FOR METALLURGICAL LITERATURE 


123 


>> u g £ 
t 05 .2 © 

| s.S’5 

JC 0> fc- © 

H PS O 

ci co 


o5 -G ^ 

« CO to 


o « t 5* O' 0 S.S g *5 

-0*5 O 2 «o « •§ 2 Sop fl 05.2 g T3 c 

ag &) CU O « C 3 5 *S q oj flj C 22 * 3 } .2 © o 6 

HxBJSSDahOQEs^ MfeOPSiSOwccftJ 

ga'asasfcsssgiJtiasgisaaa 


* 5 «5. 

»T3®’S'gS§g 
Ti © AO is 

► ~ £ g § g. 

£ « £ £ <§ o 1 5 ' 


•§ 3.1*8 

g "S |.« 
ftjj ^ -3 
co EC O fit 


® § 

o g 

J3 .5 

ft 73 

•2 c 

T3 .O * 

2 n fl 

73 g °.2 

js 8 £>-3 

6 © - 6£> « 

O » 03 © 

OS S, a « 2 fcb-JS 

a as •*» .S' a St J5 

W > £ Cl, (£ & fc 


as 

60 co 

DQ m Q) 

O c ^ 


W ^ «5 « X 05 


iO«NX050hN^NX050hN 


G c 

o S ^ 
fc: © *£ 

5*33 

g & S 2 

•c Jg g£ 
o -g a . 

«M 22 

o c o S 

si ^ ft 
t) ® 
05 05 03 

t.sic^ 


3> 1 2 3 

fl o S B 

T3 ♦a *2 

3 o g S 
« S 5 « 


W QQ W PQ 


5*3 W 5 w »a W 

» - « g *3 .2 ^ 

- # g E “ « ® 

1 S § 8. | £ I 

■ » 5 e ,i £ « o 

05 > E S G ~ fa 

ft C ® Q5 ® ft 

Q £ ft o5 o a « 




050hW«^»0«MNX 


roose powders rosion) 13. Geography 




14. Slates of Metals 21. A.PJ. 12. Atomic 

9. Minerals 22. A.M.S . 13. Solar 


124 


PUNCHED CARDS 


© 

© 


a 

© 


© 

© 

k 

© 

© 

© 

© 

si 

© 


<3 8, 


© 


© 

fe 

si 

C5 


«9 k' 96 

* « o 


© 

© 

»> § -3 
^ © "© 


© 

si 


II 


© 

© 

§•£ 


a C .2 « .h S 

Is!.S 3| 


© 

© 


© © 


5* © © g 0.-0 
© © > J2 © 3 


© 


H w CO sf »C O 


o. 

^ © 


00 


© 

© 


© 

© 

©> 

V. 

O 


© 

© 

© 


8 


© 

© 


SI 

31 

S$S 


•8 


ill 


-2 J8 


•© 

© 

•«** 

■e&„ 
oa . . . 
"T* ^ ^ 

cq <; &j 

oi Q d 


c 
o 

'S 

S § 

© 

£ £ 
i °5 
**• 6 
^ <3 


oo 

© 

© 


k. 

3 

o 


2 - 2 * 
■g 'C 

is 
$« 
* ^ 


ifl®N«oiOH.5eiOM 
h h ki) ^ 

CD 


a 

© 


"© 

©J 

"© 


© 

© 

!& 

© 

© 

o. 

CO 



«. „ .. ! H H 
| ■§ s -2 °0 60 °o 

O X Si » t ^ 

• • © © • • 

oo o> ^ t^. 00 

T—! *—I 


ft} OQ 
^ s 

CQ 


00 05 Q 

I 


»o 




Chapter 6 


THE PEEK-A-BOO SYSTEM—OPTICAL 
COINCIDENCE SUBJECT CARDS IN 
INFORMATION SEARCHING* 


W. A. Wildhack and Joshua Stern 
National Bureau of Standards, Washington, D. C. 

Introduction 

The “Peek-a-boo” principle, which is treated in detail in this chapter, 
is illustrated in Figure 6-1. The perforated cards shown represent subject 
headings or index terms. If a document being indexed relates to a particu¬ 
lar index term, the card representing that term contains a hole at a position 
dedicated to that document. Thus each card representing terms relating to 
the document will have a hole at the identical position for that document; 
all other cards will have no hole at that position. The collection of holes on 
a given card identify directly all documents relating to the corresponding 
subject. If two or more cards are superimposed, holes which are not ob¬ 
scured, i.e., those which appear on all of the superimposed cards, identify 
documents each of which relate to all of the subjects represented. The 
document serial number is represented by the location of the hole, usually 
read by means of an imprinted pattern on the card or with a transparent 
overlay. A detailed description of the mechanical operation of the system 
will be found on pages 136-141. 

The “Peek-a-boo” search mechanism was designed for a specific applica¬ 
tion and the features of the mechanism were dictated by the needs of that 
application 1 . Many other ways of applying the principle have been devised 
which are advantageous under given circumstances. The principle itself 
appears to be quite old. The earliest application known by the authors was 
for identification of birds and was the basis of a patent issued in 1915.* 
Shortly thereafter, application to a number-guessing game was described 
in a publication of very limited distribution 3 • 4 . Identification of minerals 

* Copyright excluded. 

1 Wildhack, W. A., Stern, J., and Smith, J., “Documentation in Instrumentation,” 
Am. Doc., 5, 223-37, Oct. 1954. 

* Taylor, H., Selective device, U. S. Patent 1,165,465, December 28, 1915. 

* Gerardin, A., Sphinx-Oedipe, 11, 68-70 (1916). 

4 Kraitchich, Maurice, Mathematique des jeus ou recreations mathematiques, 


125 



126 


PUNCHED CARDS 



Figure 6-1. “Peek-a-boo” system, principle of operation. (Light beams are pass¬ 
ing from left to right.) 

by application of the principle is described by Gray 5 and by Donnay* • 7 . An 
elaboration of the principle, disclosed in a patent issued in 1920 s , deals 
with means for avoiding the need to remove cards from the file for reading. 
As described in the patent, this is accomplished by interspersing fully 
perforated reading columns between the normal data-punched columns. To 
read the file, the subject cards to be compared are displaced with respect 
to the bulk of the file to cause the data columns to coincide with the inter¬ 
mediate, fully-perforated columns of the rest of the cards. All the data 
positions on the cards being compared are thus exposed to view through 
the perforations in the reading columns, unobstructed by the cards not 
involved in the search. This patent cited no prior art and the patent office 
actions did not disclose prior art. Similar systems with addition of photo¬ 
electric read out, for use in telephony, are described by Myers 9 and by 

Ed. 1, Brussels, 1930; Ed. 2, Brussels, 1953. Revised editions in English, Ed. 1, New 
York, 1942; Ed. 2, New York, 1952, London 1943. 

4 Gray, C. J., A new method of using the physical characteristics of minerals for 
their identification, Trans. Geol. Soc. S. Afric, 23, 114-117 (1920). 

* Donnay, J. D. H., Ann. soc. geol. Belg., 59, B 250 (1936). 

7 Donnay, H. D. H., .4m. Mineralogist, 23, 91 (1938). 

8 Soper, H., Means for compiling tabular and statistical data, U. S. Patent 
1,351,692, Aug. 31, 1920. 

9 Myers, O., Electromechanical translator, U. S. Patent 2,558,577, June 26, 1951. 



THE PEEK-A-BOO SYSTEM 


127 


Gent 10 . The prior art described above was not cited by either the inventor 
or examiner in these patents. Another photoelectric read out in which the 
various hole positions are scanned one at a time, so that only one photocell 
is required, is described by Drillick u . The principle has been applied to 
searching personnel records as described in a French patent issued in 1923 12 . 
Application of the principle to patent files has been made by Batten 13 ■ 
14 • ,6 . This application used hand-punched and hand-manipulated cards 
with a relatively low density of punching. An application of the principle 
which has been commercially available abroad was devised by Cordon- 
nier 18 ’ 17 • 18 . This system, believed to be based on the French patent cited 
above 12 , is trade-named “Selecto”. It uses in one form a card 15 x 21 cm 
with a capacity of 12,500 documents. The card is pre-printed with a grid 
identifying the hole positions so that no read-out device is needed. Holes are 
approximately 0.7-mm diameter. Cards of larger and smaller capacities 
are available. Also commercially available is a French system called 
“Sphinxo”. The Sphinxo card measures approximately x 10)4 inches 
and has a document capacity of 1000 19 . The authors have no information 
on the connection, if any, between this system and the disclosure cited 
above 3 , published in the Journal Sphinx-Oedipe. A card 8)4 x 11% inches, 
with 7000 document spaces, is manufactured in Germany under the name 
“Ekaha” 20 . Also of German manufacture are “Sichtlochkarten” 21 . These 

10 Gent, E. W., et al, Electromechanical translator, U. S. Patent 2,668,877, Febru¬ 
ary 9, 1954. 

11 Drillick, J. H., Fast access to punched card data file, Product Eng., 22, 176-178, 
Oct. 1951. 

12 Lieber, Henri, French Patent 565,745, Nov. 27, 1923. 

11 Batten, W. E., A punched card system of indexing to meet special requirements. 
Report of the 22nd Conference, ASLIB (4 Queens Gate, London, W. 8 ) pp. 37-39 
(1947). 

14 Batten, W. E., Specialized files for Patent Searching, Casey, R. S. and Perry, 
J. W., “Punched Cards, Their Application to Science and Industry,” Chapter 10 , 
pp. 169-181, New York, Reinhold Publishing Corp., 1951. 

14 DeGorter, B., The indexing of British patents on plastics. I.C.I. Ltd., Plastics 
Div., Intelligence Section, (C4/BdeG/JH) 18.2.47 (Hollerith). 

18 Cordonnier, G., Classification, classement, rangement et selection Revue Men- 
suelle de 1’organisation, Comite National de L’Organisation Francais, 57 Rue de 
Babylone, Paris 7e, Avril-Juillet 1951. 

17 DeMamantoff, N., Chef du Service, Le Centre de Documentation du Centre 
Nationale de La Recherche Scientifique, Utilisation du Systeme Cordonnier. (1953) 
(16 rue Pierre-Curie, Paris 5e). 

u Schurmeyer, Dr. Walther, Selecto—Ein neues Auswahlsystem fur Dokumenta- 
tion, Nachrichten fur Dokumentation, 3, * 1 , 33, Mar. 1952. 

19 Detectri, 68 rue de Richelieu, Paris 2e, France. 

* # Edler & Krische, Hannover, Germany. Ekaha cards (7000 docs. 8)4 x \1% 
inch). 

21 Allform Buro-organization GmbH, Brandenburgische Strasse 27, Berlin W 15. 



128 


PUNCHED CARDS 



Figure 6-2. “Vicref” card for hand punching. Capacity 500 documents. 


are available in approximately 5x8 inch size with capacities of 2000 and 
0000. Application of this card and of the Ekaha cards to pharmaceutical 
indexes has been described in Reference 22. 

Accounting-machine cards, and the punches available for such cards, 
have been used in applying the Peek-a-boo principle. Robinson 23 has used 
such cards to prepare stencils for solving mathematical equations. An 
application which uses a specially printed card has been devised by Seely 24 , 
and named “Vicref”—for Visual Gross Reference. The card,* illustrated in 
Figure 6-2, is pre-punched with pilot holes and pre-printed with document 
numbers. When used with a suitable hand punch, the pilot holes ensure 
perfect registry of the larger document-identification holes. The card has 
a capacity of 500 documents. Additional sets are envisaged to increase the 
file to multiples of this number and provision is made for identification of 
up to 40 sets by a series of hole positions reserved for this purpose. The 

11 Adler, Fritz H., Pharm. Ind., 19, 170-172 (1957). Aliform and Ekaba cards in 
pharmaceutical indexes. 

13 Robinson, R. M., Stencils for solving x s = 2(mod m), U. of Cal. press 1940. 

* Obtainable from Sperry-Rand Corporation. 


























THE PEEK-A-BOO SYSTEM 


129 



Figure 6-3. Samas card for hand punching, capacity 400 documents. 

authors have experimented also with the smaller accounting-machine 
cards* shown in Figure 6-3a, using an ordinary “conductor’s” punch to 
make the perforations. Recently introduced commercially in this country 
are “Termatrex” cards and equipment in two capacities, 15,000 and 40,000 
documents 25 . More elaborate implementations of the Peek-a-boo principle 
have also been made. One of these involves keyboard selection of the cards 
to be punched or read, simultaneous drilling of all selected cards, and 
random return to the file 26 . 

The information indexing and searching system to be described here was 
developed at the National Bureau of Standards as part of a program of 
research in methods of measurement and in improvement of scientific 
instruments. 

* Obtainable from Underwood Corporation. 

24 Seely, J. S., Vicref (visual cross reference), private communication 1952. 

24 Documentation Incorporated, 2521 Connecticut Ave., N. W., Washington 8, 
D. C. 

24 Miller, Eugene, Final Report to the National Science Foundation on the Matrex 
Indexing Machine, Jan. 1957, 13pp. 









130 


PUNCHED CARDS 


The program envisaged three kinds of activity: (1) research and devel¬ 
opment in instrumentation; (2) critical surveys of various areas of instru¬ 
mentation; and (3) consultation on instrumentation problems. In planning 
the program it was evident that all three activities would rely heavily on 
documentation, defined as the handling of recorded information. It was 
also evident that instrumentation literature is poorly organized and is 
hard to search, being found as a part of all branches of science and tech¬ 
nology. Studies toward improving the documentation of instrumentation 
literature were therefore undertaken involving the following elements: 
locating sources, selecting references, identifying references, indexing, 
coding, filing, storing, searching, and retrieving. These studies of the general 
problems, initiated in 1950, revealed the difficulties of conventional hier¬ 
archical classifications as the basis for indexing, and led to recognition of 
the essential “multi-dimentional” or “multiple-independent-aspect” nature 
of information, and thus to consideration of the “multi-dimensional” or 
“multi-aspect” approach to the indexing problem. 

The information indexing system which finally evolved is based on multi- 
aspect classification of document content and Peek-a-boo searching. 

Fundamental Considerations 

The key element in establishing an information file is the indexing proc¬ 
ess. This should result in the assignment of class designations which function 
to disclose to the searcher those documents in which he may expect to find 
the information he is seeking. Logical steps in the indexing operation are: 

(1) establishment of index classes to which a document may be said to be¬ 
long on the basis of its content; (2) assignment of names, or class designa¬ 
tions to the document classes thus established; (3) recognition of that con¬ 
tent information which is to be represented in the index; (4) assignment of 
the document to all appropriate index classes on the basis of content so 
recognized. A brief review of the general problems of indexing will provide 
background for a description of the system adopted for indexing, specifi¬ 
cally as applied to the literature of instrumentation. 

The following may be set forth as the specifications to be met by an in¬ 
dexing procedure: 

(1) The index must be capable of representing all information which is 
to be retrievable in the future. 

(2) The “language” of the index needs to be common to both the in¬ 
dexer and the searcher. 

(3) The classes of documents established by the index must be small 
enough to be acceptable to the searcher. 

(4) All classes of documents required by the searcher must be present 
in the index either specifically or as obvious members of more generic classes 
which are themselves small enough to be acceptable as the search product. 



THE PEEK-A-BOO SYSTEM 


131 


Each of these specifications implies anticipation of the desire of the 
searcher. To meet the first specification, the capability of representing all 
information which is to be retrievable, the index must have terms which 
subsume, or are related (logically or by arbitrary assignments) to all the 
types of information to be retrieved. The ability of the indexer to recog¬ 
nize all such information is assumed, since no index can compensate for 
failure on this point. 

A large fraction of the problems of indexing relates to the second specifi¬ 
cation, that of providing an index language common to indexer and searcher. 
The ambiguity of language in general is a basic difficulty for which a remedy 
is yet to be found. A first step is the rigorous definition of all terms used in 
the index. Such definitions may well be arbitrary, provided they are un¬ 
ambiguously understood by both indexer and searcher. More specific are 
the problem of synonyms, and the problem of generic relationships. In addi¬ 
tion to the ordinary synonym, for example, “aerial” and “antenna,” syn¬ 
onymous expressions and word-order synonyms must be recognized. Ex¬ 
amples of the synonymous expression are “force” and “rate of change of 
momentum,” and “pressure” and “force per unit area.” Examples of the 
word-order synonym are “electric motors” and “motors, electric,” or 
“radar for air navigation” and “air navigation by radar.” The problems 
presented by the word-order synonym and by the synonymous expression 
are rapidly compounded as the number of words in the index subject head¬ 
ing increases. 

The problem of generic relationships among index terms is related to the 
problem of synonyms. The expression which is synonymous with a single 
word may be regarded as a special case of generic relationships. Since the 
whole is the sum of its parts, the existence of generic relationships offers 
alternative, and hence ambiguous, use of the index terms in searching (i.e., 
a given piece of information may be sought via a general index term which 
describes it or via the combination of a number of terms which together de¬ 
scribe the information). This aspect of the generic relationship is classed 
here as a language problem since its existence results in an uncertainty in 
the mind of the searcher in regard to the choice of terms made by the in¬ 
dexer. Aspects of the generic relationship problem which influence the 
ability of the index to meet specifications (3) and (4) are treated below. 

These language problems have traditionally been treated in the follow¬ 
ing way: 

(a) The index language is formalized so that the index vocabulary is 
strictly limited. 

(b) Synonyms are eliminated or cross referenced, synonymous expres¬ 
sions being treated in the same way as simple synonyms. 

(c) Generic relationships are prevented from causing trouble by requir¬ 
ing that only established terms or established combinations of terms be 



132 


PUNCHED CARDS 


used. In this way it is possible to provide a fixed hierarchy so that all recog¬ 
nized generic relationships are explicit, permitting the searcher to select 
the most specific index term which meets his needs. 

(d) The word-order synonym is treated by cross reference in the same 
manner as are other synonyms. No more than a minute fraction of the 
problem can be dealt with in this way because of the tremendous number of 
such synonyms which exist. 

These remedies for the language problem are limited severely by the 
ambiguity of language which makes impossible the establishment of a 
universally understood index vocabulary which will not be eroded by time. 
Furthermore, while the techniques mentioned contribute in good measure 
to the solution of the language difficulty, they are not entirely compatible 
with the other requirements of indexing, particularly specifications (3) 
and (4). 

Specifications (3) and (4) deal with establishment of the classes of docu¬ 
ments which are to be the product of the information search. These require¬ 
ments of an index are basic. The total collection (representing the most 
general class of documents) must be broken down into classes small enough 
to be searched further, without additional help from the index. (The exist¬ 
ence of auxiliary specific indexes influences the acceptable size of the product 
class. For example, if the document is a book, its own index will carry the 
burden of searching within the document.) The size of the search product 
which is acceptable varies widely with the circumstances of the search so 
that it is not possible to correlate the kind of information to any “accept¬ 
able” size. Usually arbitrary bounds are set on the kind of service to be per¬ 
formed by a given index. When one is seeking a specific datum for imme¬ 
diate application in conjunction with other data only the most specific 
class will be acceptable. For example, in seeking information on the melting 
point of a particular organic compound, a class which contained all docu¬ 
ments in which properties of organic compounds were discussed would be 
very unsatisfactory. For this kind of search a handbook type of index system 
which relates the specific property and the specific compound is necessary. 
On the other hand, a search designed to disclose the state of the art of 
measurement of flow of liquids would tolerate—or, in fact, demand—a 
very large search product. The kind of service to be fulfilled by the index 
needs to be anticipated and carefully defined. If one considers the establish¬ 
ment of index classes without anticipating future meaningful reference 
questions, the number of valid classes which can be defined approaches in¬ 
finity. Even when limits are placed on the system by anticipating the kinds 
of questions which will be asked, it is still necessary to provide a larger 
number of search-product classes than can be recognized explicitly. At¬ 
tempts to recognize a sufficient number of meaningful specific document 



THE PEEK-A-BOO SYSTEM 


133 


classes magnify enormously all aspects of the language difficulties. The 
cross-reference technique quickly becomes inadequate. An approach to 
indexing which can provide the large number of classes needed is that which 
has been referred to above as the multi-dimentional or multi-aspect ap¬ 
proach, in which each search-product class is specified in terms of common 
membership of each document in a number of more general classes desig¬ 
nated explicitly by general index terms. The number of specific document 
classes which can be formed in this way far exceeds the number of general 
classes (index terms) which are explicitly established. For example, with a 
vocabulary of 1000 general terms the number of specific classes which can 
be generated by combining up to 10 terms in every possible way is approxi¬ 
mately 10 22 . Of course many of the classes established in this way will not 
relate to any actual document and some may include contradictory ele¬ 
ments and hence will not have even potential utility. 

The principle of combination thus gives the very large number of classes 
needed without requiring that each be separately recognized in advance. 
Whether the classes which can be defined in this fashion include the ones 
which are needed for search purposes remains to be seen. If the class names 
recognized explicitly in the index consist exclusively of single words, no 
word-order problem can arise. It is usually not desirable to do this fully. 

This approach implies abandonment of some of the techniques listed 
above which have been used in the past to control the “language” problem 
as outlined above. The principle of formation of specific classes at the time 
of search by combination of a number of general class designations is not com¬ 
patible with rigid formalization of the search language. This in turn means 
that the synonym problem is increased in severity by the variety of ways 
in which a given document can be specified by combinations of index ele¬ 
ments. These factors, particularly when combined with the ambiguities 
inescapable in any system, and with the uncertainty whether the increased 
number of classes implies a significant increase in the number of useful 
classes, raise questions as to the net gain derivable from the newer tech¬ 
niques. Nonetheless, it is possible to conclude that there is a net gain, on 
the basis of specific advantages, even though these too may not be suscepti¬ 
ble to quantitative evaluation. If we regard an index with no provision for 
generating new classes by combination of recognized classes as one extreme 
of a spectrum of index systems, and the system in which recognized classes 
are never used alone, but are always combined to form the end-product class 
sought, as the other extreme, then it is evident that all systems fall some¬ 
where within this spectrum. The problem of designing an index which uses 
combination techniques is that of finding the optimum position along this 
spectrum. This assumes that effective techniques can be found to solve the 
mechanical problem of performing the operation of combining classes to 



134 


PUNCHED CARDS 


form other classes and to isolate the resulting classes. The following discus¬ 
sion deals with these techniques and in particular with the techniques to 
which the name “Peek-a-Boo” has been given. 

Search Mechanisms 

Having devised a search classification or index which depends upon the 
common association of a document with a number of different index terms 
to identify that document, it remains to select mechanisms for recording 
and detecting such associations. A wide variety of such mechanisms have 
been devised, ranging from simple, hand-manipulated cards to the most 
complex electronic computers. Many more such devices have been proposed 
or can be conceived. 

By and large, devices for this purpose fall into two classes: those in which 
information identifying a document is associated, inseparable physically, 
with all the index information relating to the document, and those in which 
symbols for all documents relating to a given item of index information are 
associated, in a physically inseparable fashion, with each such index term. 
If the physical entity accomplishing the necessary association is a card, 
we have on the one hand “document cards” each of which carries all the 
pertinent subject headings and, on the other hand, “subject cards” each 
carrying all of the pertinent document identifications. The same situation 
holds if we substitute for the card, as the medium for associating index and 
document, any of the other means which have been used or proposed such 
as punched tape, photographic film, magnetic wire records, magnetic tape 
records, static magnetic memories, superconductive memories, etc. 

It is convenient to classify indexing systems in this way—as document- 
card type or subject-card type—because with each of the two classes are 
associated important performance characteristics which need to be •weighed 
in terms of a specific application to estimate its advantages and disadvan¬ 
tages. 

A system characteristic of prime importance is the nature of the presorts 
permitted—a sort being defined for this purpose as the physical formation 
or isolation of a class as opposed to the mere definition of that class, and a 
presort being such a sort made at the time of filing rather than at the time 
of search. No matter how fast our mechanisms can search a collection of 
cards, film frames, or magnetic states, an additional increment of speed is 
obtainable if only a portion of the collection needs to be searched. This 
principle is, of course, the controlling one in the classic hierarchical ap¬ 
proach to indexing which postulates “a place for everything, and every¬ 
thing in its place.” All presorts which are based on setting up preferred 
groupings necessarily require anticipation of the search questions which will 
be asked of the index in order to select from the vast number of possible 



THE PEEK-A-BOO SYSTEM 


135 


groupings those which are to be preferentially presorted. It is therefore a 
matter of intuition, or guesswork, to predict the nature, extent, and im¬ 
portance of future failures. In systems which use the search principle of 
combining general indexes at the time of search to generate specific indexes, 
it is possible to make one kind of presort, as defined above, without requir¬ 
ing arbitrary preference. This is not because the presort is a perfect one but 
because it adds no additional imperfection beyond that which has already 
been accepted in the system. Given a set of general indexes which are to be 
combined for more specific searching, it is always possible to presort on the 
basis of each of the general indexes without introducing any new ambiguity 
into the future search. The reason for this lies quite simply in the fact that, 
in any case, every search must start by specifying at least one index. The 
presort on the basis of each index is the identical operation which must in¬ 
evitably constitute the first step in the search. If such a presort has been 
made, then instead of searching the entire collection, it is only necessary 
to search further those items which have already been grouped in a com¬ 
pact form. The resulting saving of search time can be decisive. The nature 
of this presort will become clearer when a particular system is described. 

This susceptibility to an unambiguous presort of this kind is characteris¬ 
tic of the class of search systems in which a medium, such as a card, is used 
to associate each index term with all the documents to which it pertains. 
For convenience we will henceforth refer to this class as the subject card 
class and to its converse as the document-card class, keeping in mind that 
the considerations involved are not limited to card systems. 

A second characteristic, relatively less important, is the ability of a sys¬ 
tem to yield subject information as a direct product of a search—rather 
than a set of serial-number citations. This capability is a direct characteris¬ 
tic of the document-card class. It is possible to make a subject-card system 
perform in this manner as will be shown later. 

Closely connected to the preceding is another characteristic relating to 
the adaptability of a system to replication and to current distribution of the 
information “cards.” Since indexing is perforce a document-by-document 
operation, the document-card system impresses no constraint on current 
distribution (analogous to the unambiguous subject presort of subject 
cards). The subject-card file, on the other hand, must await filling of the 
card before distribution of duplicates can be effected economically. It is 
feasible to distribute annual indexes in this way, replacing an incomplete 
set with an up-to-date one periodically. Alternatively, an initially com¬ 
plete index could be distributed, followed by currently distributed docu¬ 
ment analyses which the recipient would “punch in” to keep the index cur¬ 
rent. 

Of great importance is the “open endedness” of the file for additions to, 



136 


PUNCHED CARDS 


and modifications of, the index vocabulary and for growth of the collection 
of documents. Document-card systems are limited in open endedness 
with respect to subject. A document-card system therefore demands very 
careful design of the subject index. Modifications tend to disrupt the sys¬ 
tem. Additions are usually severely limited by the capacity of the medium 
(card). The subject-card system has no such limitations, the addition of 
another subject index term requiring only an additional card. On the other 
hand, the subject card must necessarily have a limited capacity for 
document codes. Limitations on open endedness of either system are not 
absolute. In either system additional index terms in one case, or additional 
documents in the other, may be accommodated by using multiple cards. 
Additional cards used in this way are called “trailer cards.” The use of 
trailer document cards involves the difficulty that selection may require 
interaction of subjects which appear on separate cards. This is not easily 
accomplished. Trailer subject cards do not encounter this difficulty since 
it is never necessary to compare the information on a card with information 
on any of its trailers. 

In selecting a subject-card system rather than a document-card system 
as the search mechanism for use in the multi-aspect instrument reference 
system at NBS, the overriding considerations were: (1) the desirability of 
subject open-endedness for an experimental system, (2) the mechanical 
advantage in rapidity of search provided by the subject presort characteris¬ 
tic of the subject-card system. After some initial experimentation with sub¬ 
ject cards on which were listed the serial numbers of pertinent documents, 
the subject-card system based on positional coincidence of perforations 
was selected as the most effective, mechanically simple, subject-card system 
known to us. 

The Peek-a-Boo System. The name Peek-a-Boo was given to the par¬ 
ticular subject-card system used for the instrumentation reference file as 
being noncommittal as to origin and descriptive of the operating principle 
which appears to have a rather long history (page 125), dating back at 
least as far as 1915 2 . 

The two cards shown in 6-1 represent index cards, with each one repre¬ 
senting an index term. The position of each and every hole in a card is 
interpretable as a serial number of a corresponding document. If several 
such cards are superimposed, only those hole positions which are perforated 
in all of the cards will permit light to pass through. Hence, such illuminated 
positions identify documents with which all of the index terms represented 
by the superimposed cards are associated. 

In adapting this principle to the instrument reference file, a card is as¬ 
signed to each descriptive term in the index. The descriptive term is noted 
at the top of the card and the cards are filed alphabetically within each 



THE PEEK-A-BOO SYSTEM 


137 


category or sub-category. As an example, a document on “electromagnetic 
flowmeters for blood” would be recorded in the file on one card labeled 
“electromagnetic,” on one labeled “flow,” and on another labeled “blood” 
(or “liquid,” if liquids were not further subdivided). The single card labeled 
“flow” would refer not to the one report of the example (Serial No. 10,924) 
but also to all other reports in the collection which have the word “flow” as 
one of their descriptive terms. The descriptive term card directs the 
searcher to the reference by means of the visible hole, one for each report 
to which that descriptive index term has been assigned. The location of 
this hole (row and column) identifies the serial number, and possibly the 
location, of the report. Thus the card labeled “flow” contains a hole in 
position 10,924; the card labeled “electromagnetic” also contains a hole 
in position 10,924, and the card “blood,” or “liquid” also carries a hole in 
that position. 

The Peek-a-boo instrumentation file is based on a card measuring ap¬ 
proximately 5x8 inches with a capacity of approximately 18,000 ordered 
spaces for holes. 

The Card. The material used for the card is a matter of some importance. 
Durability, dimensional stability, opacity, suitability for typewriting, 
punchability, permanence, and availability are factors which enter into 
the choice of material. Paper, which was used in the initial experimentation, 
was found deficient in durability and dimensional stability. This, however, 
is of importance only when a high density of information storage is desired. 
It is possible by appropriate design to overcome the difficulties resulting 
from dimensional instability 1 if one is willing to accept some operating 
limitations. 

The Peek-a-boo cards more recently used at NBS are of vinylite plastic.* 
They are negligibly affected by humidity. Those qualities which require a 
longer time for their complete evaluation appear to be satisfactory so far. 
Typing requires a special ribbon which is, however, inexpensive and readily 
available. One corner is cut to ensure unambiguous orientation. Except for 
the index terms typed in the upper left corner, there are no markings on the 
card. Cards imprinted with a grid which would permit document numbers 
to be read directly were not obtainable at an acceptable price in the small 
quantities required. The need for precise printing registry and special card 
stock contribute to the high cost of preprinted cards. Figure 6-4 illustrates 
the cards and shows how the juxtaposition of two cards narrows a search 
based on two general terms. 

Punching. The punch used to perforate a hole in the proper position 
corresponding to the serial number identifying a document is shown in 
Figure 6-5. It is designed to have sufficient precision to take advantage of 

* Obtainable from Wassell Organization, Inc., Westport, Conn. 



138 


PUNCHED CARDS 


r-— ^ 



Figure 6-4. a) Peek-a-boo card showing documents relating to measurement of 
flow, b) Peek-a-boo card showing documents relating to electromagnetic principle 
of measurement, c) Superposition of the general “flow” card and the general “elec¬ 
tromagnetic” card to specify documents relating to electromagnetic measurement of 
flow. 


the dimensional stability of the card. The punch perforates holes approxi¬ 
mately 0.6 mm in diameter in a square pattern spaced 1 mm center-to- 
center. In this way, an area 100 mm x 180 mm (4 x7 ft inches approx.) on 
the 5x8 inch card accommodates 18,000 documents. Positioning is accom¬ 
plished by two screws, one for each coordinate of motion. A single turn of a 
screw moves a fence one unit horizontally or vertically. The card to be 
punched is inserted so that two edges are located by the fence. Thus the 




THE PEEK-A-BOO SYSTEM 


139 



Figure 6-5. Index punch for Peek-a-boo cards. 


position for successive documents is varied with respect to a fixed punch 
and die. The normal procedure in punching information is to proceed in the 
order of document serial number. In this procedure, adjustment of the 
punch to the proper position for successive documents is made by one turn 
of a knob. A detent eliminates any concern with visual setting of the knob 
by providing a “feel” for the correct position. If it is necessary to backtrack 
or to punch documents out of order, a split-nut release mechanism permits 
rapid traverse to the desired positions. Scales on the two axes identify the 
setting of the punch and provide a useful check from time to time. 

Punching errors, when they are discovered, can be corrected in a simple 
manner. The punch is set to the position at which the incorrect hole is 
located, and the incorrectly punched card is placed as if for punching. A 
piece of paper card stock, of ordinary index-card thickness, is placed over the 
plastic card and the perforating die is blocked by placing a suitable piece of 
metal shim stock under the card. The lever is then depressed, punching the 
paper card and forcing the paper punching into the hole in the plastic card. 
No cementing appears necessary; the fibrous insert spreads sufficiently to 
lock itself firmly in place. 


140 


PUNCHED CARDS 


The index designed for the field of instrumentation includes approxi¬ 
mately 1000 primary terms (i.e., terms represented by cards). These terms 
are typed on the Peek-a-boo cards with a preceding number to identify the 
category (see Figure 6-4). The cards are filed by category and by alphabet 
within categories. Information to be punched is furnished to the operator 
in the form of 3 x 5 inch abstract or citation cards, on the reverse of which 
are the index terms assigned by the document analyst. These “master” 
cards are serially numbered and are presented to the operator in serial order. 
Assuming the operator has just completed punching for a prior document, 
the punch is adjusted to the next position by one rotation of the “units” 
knob. Each pertinent Peek-a-boo card, selected from the file in turn as 
called for by the master card, is inserted into the punch, punched and re¬ 
turned to the file. When all the index terms have been thus recorded, the 
punch is set to readiness for the next document by rotating the indexing 
knob. When a column of 100 hole positions has been completed (100 docu¬ 
ments entered), the fence carriage is returned to the units “zero” position 
by releasing the split nut on the “units” screw. The “hundreds” screw is 
then rotated one turn advancing the carriage to the next column, and punch¬ 
ing is resumed, document-by-document. 

Experience thus far demonstrates a punching rate of 2000 holes per day 
(approximately 200 documents) for a single machine. This rate allocates 
an acceptable fraction of total cost to the punching operation. Enhance¬ 
ment of the rate is expected to result from a modified arrangement of the 
Peek-a-boo file which will stagger the cards in the file to make the headings 
directly visible, as is done for example with Ekaha cards 1 *. 

When all of the hole locations on a set of cards have been utilized (i.e., 
18,000 documents) it is necessary to establish a new set. For the rate of 
accumulation anticipated to cover the literature of instrumentation this 
represents roughly one set per year. For the punching operation only one 
set is active at any one time so that the addition of sets has no influence on 
this operation. The effect on the searching process is discussed below. 

Searching. Since the card bears no imprinted grid from which the posi¬ 
tion of holes may be read, an overlay, consisting of a transparent sheet 
ruled with a millimeter grid with appropriately numbered scales, is used. 
Readout may be accomplished by holding the stacked cards, together with 
the overlay, in proper register against a bright background. The overlay 
superimposed on cards to be read, stacked together on a transparent plastic 
rack, is shown in Figure 6-6. 

The search for information starts with a “translation” of the search ques¬ 
tion in terms of the index vocabulary. The selected terms may be grouped 
in various combinations of different specificities as a guide to the searcher. 
This is particularly desirable if the actual search is to be performed by a 



THE PEEK-A-BOO SYSTEM 


141 



Figure 6-6. Read-out stand for Peek-a-boo cards. 

clerk. Instructions to the searcher may request, for example, that restric¬ 
tive terms be added in a given order until the number of documents dis¬ 
closed has been suitably limited. The searcher withdraws the requisite 
index cards from the file, places them on the rack together with the over¬ 
lay, and notes the document numbers indicated by the intersection of 
vertical (hundreds) and horizontal (unit) grid lines. It has been found con¬ 
venient to have the holes fall within squares and to establish the document 
number by taking first the three digits of the line to the left of the hole and 
then the two digits of the line below the hole. Entering the “master” file 
with these document numbers, the searcher obtains the corresponding cita¬ 
tions and abstracts. 

Refiling may be facilitated and errors in filing quickly found and cor¬ 
rected if a diagonal stripe or groove is marked on the upper surface of the 
collection of cards as seen in the file drawer. 

Application to Instrumentation Literature 

In the preceding pages we have discussed an indexing approach based, at 
least in part, on a search technique which requires that a number of general 
classes be combined to produce a more specific class. An effective mechanism 
for carrying out this operation has also been described. It remains to dis¬ 
cuss the procedure for setting up such an index and to describe some of the 




142 


PUNCHED CARDS 


techniques which permit maximum benefit to be derived from the multi- 
aspect approach and minimum loss to be suffered from its deficiencies. This 
will now be done by describing the application of multi-aspect indexing and 
Peek-a-boo searching to instrumentation files maintained at the National 
Bureau of Standards. While this is a specific application with some prob¬ 
lems and features peculiar to itself, most of the problems and requirements 
dealt with in setting up the index are quite general and the ways of 
dealing with them are immediately applicable in other documentation areas. 

Broadly, the vocabulary established for the instrumentation indexing 
system is based on ten categories which represent major “points of view” 
from which the information seeker is expected to approach the search. To 
these categories are assigned two kinds of terms, the index terms proper 
(primary terms) and other terms (synonym or referred terms) which refer 
the searcher to. appropriate index terms. The document searcher, analyst 
or indexer assigns terms from the categories to the document without being 
restricted in any way by the grouping into categories. He may use terms 
from one or more of the categories as appropriate and any number of terms, 
or none, from a given category. 

The ten categories designated for the field of instrumentation and the 
approximate number of primary terms in each, are: 

(1) Measurands (100 primary terms). These are the terms representing 
the variable, quality, quantity, or condition, the measurement of which is 
the subject of the document. Typical terms are: acceleration, force, dis¬ 
placement, brightness, magnetism. 

(2) Principles (300 primary terms). These are the terms which denote the 
operating principles of the instruments or methods by which the measure¬ 
ment is performed. Examples are electromagnetic, sonic, radiation, spectra, 
resonance. 

(3) Object (100 primary terms). The materials or things of which the 
measurand is a property, for example, copper, structure, fluid, celestial 
body, vehicle. 

(4) Instrument Name. This category contains no primary index terms, 
only referred terms, because no way has been found to generalize the large 
number of special names without redundancy with other categories. For 
example, if the measurand is voltage, the instrument is likely to be a volt¬ 
meter, if its operating principle is electrostatic the instrument is more 
specifically an electrostatic voltmeter. The very specific individual instru¬ 
ment names, particularly trade names of instruments, are listed with defi¬ 
nitions which reveal the measurand and operating principle, thus referring 
the searcher to the appropriate index terms in other categories. 

(5) Field of Application (60 primary terms). These terms indicate the 
area of science or technology in which the document originated. This is 



THE PEEK-A-BOO SYSTEM 


143 


frequently quite different from the field defined by the measurand itself 
and offers a powerful tool for refining the search. For example, if one is con¬ 
cerned with the measurement of acceleration, the fact that a paper on this 
subject originated in connection with aviation may have a great bearing 
on the likelihood that the information contained therein is appropriate to 
the needs of the search. Examples of other fields typically represented in 
the index are: instrumentation itself, research, medicine, navigation, 
process industry. 

(6) Development Stage (10 primary terms). In the course of develop¬ 
ment of an instrument various stages are recognizable and papers appear 
which deal in major aspect with particular stages. Examples of the index 
terms to be found in this category are: design, construction, prototype, 
production, testing, calibration. 

(7) Function in Instrumentation Process (50 primary terms). The 
process of measurement is conceived as consisting of a series of functions. 
Examples are detection, amplification, transmission, display, recording, 
computation, control. These functions may form a sequence of operations 
in a given case although no fixed sequence can be assigned to the functions 
in general. Frequently specific functions are a major subject of a document 
which can be recognized by index terms drawn from this category. 

(8) Character of Document (50 primary terms). Such descriptive charac¬ 
teristics as language, government report, bibliography, book, history, re¬ 
view, are indicated here. Like category (5) this category is quite generally 
applicable. 

(9) Characteristics of Device (50 primary terms). Here are indexed such 
performance parameters as accuracy, precision, ranges, environmental 
limitations and such features as portability, etc. 

(10) Limitations on Measurand (200 primary terms). This category is 
designed to cooperate with category (1) to specify the measurand further. 
Many of the index terms in this category are duplicates of terms in (1). The 
fact that they are in category 10 gives them a different significance. The 
word “temperature” in category (1) signifies that the document deals with 
the measurement of temperature. The same term in category (10) signifies 
that the measurement in question is somehow limited by temperature con¬ 
siderations; for example it may be concerned with measurement of pressure 
at high temperatures. 

Categorization is expected to provide a number of advantages. First 
among these is guidance of the thoughts of the indexer and the searcher 
along similar tracks. Related to this is the added assurance that the indexer 
(and the searcher) will give consideration to all of the pertinent aspects of 
documents relating to instrumentation, although it is not expected that 
every document will have a term from every category. The categories also 



144 


PUNCHED CARDS 


serve to distinguish between identical words used with different sense, as in 
the example cited of the significance of the word “temperature” as a 
measurand in cateogry (1) and as a limitation in category (10). The use of 
categories is a recognition of the fact that it is possible, in any information 
area, to select at least a few groupings of related terms which are universally 
acceptable and may be expected to remain stable. In making use of 
relationship of the categories to the terms subsumed within the category, 
advantage is being taken of desirable features of conventional indexing. 

The Index Terms. The indexing terms, conveniently referred to as pri¬ 
mary terms to distinguish them from the much larger number of secondary 
terms (which are cross referenced to the primaries and which are not them¬ 
selves separately represented by Peek-a-boo cards), at present number about 
1000 for the instrumentation reference file. A conscious effort has been made 
to keep the number of primary terms small. No limit is placed on the num¬ 
ber of secondary terms listed in the search dictionary, approximately 2000 
at present. It is believed that these should include every term which may 
be thought of by the technical searcher and should disclose how each is 
represented in the index. 

A number of important advantages derive from a small vocabulary of 
primary index terms: 

(1) A small vocabulary implies a corresponding reduction in the severity 
of the synonymous-expression problem. The number of meaningful com¬ 
binations of terms, and hence the number of synonymous expressions of a 
class (i.e., different combinations of terms which express the same idea), 
increases rapidly with the size of the vocabulary. Each of these offers the 
indexer or searcher an alternative choice of combination terms, only one of 
which is correct. 

(2) The labor of finding cards in the alphabetical file, for punching and 
searching, is roughly proportional to the number of cards. 

(3) A small vocabulary will be learned rapidly by the indexer resulting 
in more rapid assignments and in fewer erroneous assignments. The same 
consideration holds true for searching. 

A small vocabulary is achieved at the expense of a larger dependence on 
combining terms to express specific index ideas. This increases the number 
of terms assigned to the average document, thus increasing the probability 
of false associations and the punching load. A proper balance must be 
achieved among these opposing factors. In order to avoid reintroducing the 
word-order problem, composite terms are used very sparingly. In principle, 
apart from the word-order problem, the accidental nonexistence of a single 
term to express a desirable index idea is not permitted to exclude the use of 
that idea as a primary term, since many composite terms have acquired a 
preferred word order and hence perform in all respects as single terms. It is 



THE PEEK-A-BOO SYSTEM 


145 


frequently found in devising a set of primary terms, that no word is avail¬ 
able to express correctly the content of a desired primary index class. In 
such cases it is useful to select as the index term a word which describes a 
part of the class and to use this term arbitrarily as a code which represents 
the class. The meaning of the term then is a mnemonic rather than a pre¬ 
cise indicator of subject class content. Thus the word “capacity” might 
designate a class which also includes “capacitor,” “capacitance,” and “vol¬ 
ume,” as well as “elastance,” the reciprocal of capacitance. It may, in fact, 
be useful at times to go farther in regarding the term as a “pronounceable 
code” and group items in a class for no logical reason other than usefulness 
in searching. As a result of this kind of technique some of the referrals from 
secondary terms to primary terms do violence to the accepted meanings of 
terms and may cause the purist some discomfort. 

The lists of primary terms have undergone several major revisions and 
spot revisions are expected always to be in order. 

Document Analysis. The document analyst assigns to a document 
terms selected from the index vocabulary and he may designate portions of 
the text which are to constitute an abstract. The term assignments, the 
reference citation, and the abstract, if used, are noted on a 3 x 5 inch card 
which becomes the master card for the document. This card is assigned a 
serial number which designates the location at which its representative hole 
is to be punched in the Peek-a-boo file. Indexing words which suggest them¬ 
selves to the analyst, and are not in the dictionary of terms, are noted and 
at a later date are entered into the dictionary, and assigned “see” refer¬ 
ences. The all-embracing nature of the dictionary built up in this way is 
believed to be a vital element of the reference system. 

Document analysts are technically trained in fields basic to instrumenta¬ 
tion. Experience so far does not suggest that the new approach is at all 
different from conventional indexing in the need for subject experts to 
analyze documents or in the level of technical competence required. 

Further Developments 

The versatility of the indexing system described permits a variety of 
techniques which have not yet been applied in the instrumentation refer¬ 
ence file but which are expected to prove valuable in the future or in other 
applications. Some of these will be discussed here. 

Unconventional Subject Terms. Many concepts, not usually regarded 
as subject descriptors, provide useful search indicators. These can often be 
treated in the same way as normal index terms. For example, it is usual to 
regard the author index as quite separate from the subject index. Neverthe¬ 
less, reference questions frequently arise which require interaction of author 
and subject. It is also a matter of considerable convenience to be able to 
maintain one general file instead of a number of special-purpose files. 



146 


PUNCHED CARDS 


Authors can be indexed in Peek-a-boo in the same way as subjects. To 
avoid inefficient use of cards and a great increase of the number of cards 
required, each author is not represented by a card but is alphabetically 
grouped with other authors. For example, the first three consonants in the 
name might be used to define groups requiring 3 sets of 26 alphabet cards 
for their identification. Interaction between author groups and very general 
subject classes is used to separate a particular author from others in his 
group. 

Chronology, “in-file” indications, journal titles, institution of origin, and 
country of origin may also be the basis for search cards. Numerical informa¬ 
tion may be recorded on Peek-a-boo cards by using cards to designate digits 
to any degree of fineness. Other cards may be used to record exponents of 
10. A precision of 1 per cent (2 significant figures), and a range of 20 orders 
of magnitude, may be recorded with 20 digit cards (2 sets of 10), 10 expo¬ 
nent cards, and 2 cards to identify sign of the exponent. Thus the number 
1200 may be recorded by punching a “1” card, a “0.2” card, an exponent 
“3” card, and a “+” card; and the number 0.12 by punching the same 
digit cards, an exponent “1” card, and a “—” card. Various other systems 
can be devised to exchange economy in total number of cards for economy 
of punching. 

Converse Searching. It is sometimes desirable to be able to search for 
the absence of a given characteristic. That is, in contrast to the normal 
search, one wishes to find all documents (or other entities) which have 
none of a group of index characteristics. Such a situation occurs, for exam¬ 
ple, in an application of the Peek-a-boo principles to indexing infrared 
spectra being developed by Savitzky 27 . In searching infrared spectra for 
identification of constituents of an unknown mixture of which the spectrum 
has been measured 18 , it is desired to find all compounds which have no in¬ 
frared absorption bands in any of a number of frequency ranges. It is, of 
course, possible to have “no-band” cards for this purpose. This would, 
however, require that each “no-band” card be punched in the large num¬ 
ber of cases in which no absorption band is present, imposing a very large 
punching burden. It is possible to reduce greatly the burden by inverting 
the usual punching operation. In this inversion, a transparent card, or a 
totally pre-punched card, is used instead of the usual opaque card. Then 
it is not necessary to punch the multitude of cards which do not characterize 
the document (or spectrum), only to treat the cards which do characterize 
it, by filling the perforation (on a pre-punched card) or otherwise obscuring 
a spot (on a transparent card). To minimize punching, each index card 

21 Savitzky, Abraham, Perkin-Elmer Corp. Private communication. 

18 Baker, A. W., Wright, N., Opler, A., Automatic Infrared Punched-Card Iden¬ 
tification of Mixtures. Anal, Chem., 25, 1457 (1953). 



THE PEEK-A-BOO SYSTEM 


147 


"MICROCITE" SYSTEM 
(EXPLODED VIEW) 



"MICROCITE" MATRIX / 

BIBLIOGRAPHY CARO 

Figure 6-7. “Microcite” system, principle of operation. 


could be paired with its pre-punched converse and the operation of punch¬ 
ing could simultaneously fill the corresponding converse hole (as in the 
technique of repairing errors previously described). Photographic techniques 
for preparing positive and negative copies are also obviously applicable. 

“Microcite”. In the section dealing with the comparison of subject cards 
with document cards, it was pointed out that the serial numbers yielded by 
the subject-card search constitutes an intermediate step which carries with 
it some undesirable features. To overcome this shortcoming a development 
has been undertaken, which has been named “Microcite,” the principle of 
which is illustrated in Figure 6-7. With the Peek-a-boo search set is asso¬ 
ciated a matrix film on which is photographed, with suitable reduction, the 
equivalent of the master card (3" x 5" card) used in normal Peek-a-boo 
searching (i.e., citation, title, abstract, etc.). Each of these is located on the 
matrix film in the position dedicated to the corresponding document on the 
punch cards or in a related position. All of the area of the film can be utilized 
for the photographs because, in contrast to perforations, the photographed 
areas require no supporting areas. In using Microcite to prepare bibliog¬ 
raphies in answer to reference questions, the film matrix (a negative in 
this case) is to be sandwiched between the Peek-a-boo search stock and a 




148 


PUNCHED CARDS 



Figure 6-8. Card punch for 1000 document cards used in experimental Microcite 
system. 

suitable photographic printing paper. Light passing through the Peek-a-boo 
holes provides illumination for printing, thus printing only the selected 
references. A suitable diffuser between the Peek-a-boo stack and the film 
matrix permits the small hole to illuminate the larger area of the master 
card image. 

An experimental version is based on a card with 1000 hole positions. 
Photographic reductions of 30 to 1 are used and a typewritten area of ap¬ 
proximately 3x5 inches is recorded in the matrix area devoted to each 
hole. The simple punch developed for this experiment is shown in Figure 
(i-8. The kind of bibliography card obtainable with the experimental micro¬ 
cite is illustrated in concept in Figure 6-9. 

Extension of Microcite to the main Peek-a-boo instrumentation reference 
collection will require 12 film matrices (or one larger matrix) for each Peek- 
a-boo set if no loss of master-card information or higher reduction ratios 
are assumed. With the very small hole areas involved, further development 
beyond the experimental version will be necessary to accomplish illumina¬ 
tion of the appropriate film-matrix area for photographic printing of bib¬ 
liographies. However, another use of Microcite, possibly more important 
than the preparation of microprint bibliographies, involves no such develop¬ 
ment problems. This is the use of Microcite in guiding the search process 
by providing for observation of the course of the search. For this purpose 
the Peek-a-boo holes are used only to locate the corresponding microphoto- 


THE PEEK-A-BOO SYSTEM 


149 



Figure 6-9. Concept of bibliography card product of Microcite search. 


graphed citation on the matrix and not to illuminate the area for reading- 
Reading is accomplished by reflected light, with a suitable magnifier, 
microscope or microprint reader or by projection. Construction of suitable 
equipment for this application is underway for preparation of film matrices 
for the main instrumentation file. 

Application to Large Collections. The use of Peek-a-boo cards for 
large collections involves either multiple sets or larger cards. The use of 
multiple sets is quite straightforward and appears to be feasible at least 
for a small number of sets. Experience does not as yet permit a limiting 
number to be estimated. Great increases in card size appear also to be 
feasible. For example, a card approximately 42 x 22 inches, accommodating 
500,000 documents, does not appear to be unreasonable for manipulation 
in searching or for the demands made on dimensional stability of cards. 
Manual punching would probably be abandoned in using such cards. Other 
reasons exist for giving serious thought to further mechanization of the 
punching operation. Punching as it is now performed is a fairly costly 
operation (though minor as compared with document analysis), and adds 
complexity to production scheduling. 



150 


PUNCHED CARDS 


A revised operating procedure, now in the design stage, promises to im¬ 
prove punching efficiency and to permit extension to much larger cards. 
This procedure is based on the use of preliminary Peek-a-boo sets, to be 
punched directly by the document analyst. This will replace handwritten 
notation by the analyst and should not consume additional time. In any 
event, the analyst’s work is paced by the thinking necessary for effective 
analysis and not by the writing or other mechanical operation he needs to 
perform. An intermediate card of 1000-document capacity appears suit¬ 
able and it is planned to use the Microcite punch (Figure 6-8) for this pur¬ 
pose. The intermediate sets of the various analysts are to be collected at 
intervals and the information of these sets transferred automatically to 
larger cards with greater density of hole positions. The transfer punch used 
to accomplish this is still to be constructed. A number of advantages are seen 
in this modus operandi beyond those which initiated its consideration. 
It lends itself very well, for example, to a “contributing analyst” organiza¬ 
tion of effort. The punch is simple and efficient in operation; intermediate 
card sets can be imprinted with index headings by automatic machinery; 
the individual sets, identified by assigned blocks of numbers, become sub¬ 
collections of special utility to the contributor; the sets can be combined 
in a variety of ways with other sets, not necessarily the entire collection, 
for special purposes; the subsets provide a means for production of repli¬ 
cate main sets for distribution or publication. 

Very Large Collections. It is not possible at this point to guess at the 
maximum size of collection for which the visual Peek-a-boo technique is 
effective, or even to predict whether the Peek-a-boo technique or the multi¬ 
aspect classification itself will be first to break down with increasing size. 
It is useful, nevertheless, to consider the characteristics of electronic sys¬ 
tems analogous to Peek-a-boo to see whether these may be expected to 
handle very large collections which are becoming of more and more common 
concern. The basic advantages of subject card versus document card are 
certainly not limited to the punched-card medium. The disadvantages of 
subject cards are in fact less significant when the concept is extended to 
more automatic operation. There would certainly be less occasion to dis¬ 
tribute very large systems than smaller ones and increasing document 
capacity implies only an increase in size and not a less convenient operation. 
A number of electronic analogs of Peek-a-boo have been outlined of which 
the following is an example. 

On a multitude of magnetic-recording channels (magnetic wire, tape or 
drums may be employed for example) are to be recorded “bits” or signals 
(pulses) whose positions along the channels represent document serial 
numbers. The individual channels themselves then are analogous to Peek- 
a-boo cards and represent index terms. The serial-number “pulses” might 



THE PEEK-A-BOO SYSTEM 


151 


be recorded as amplitude modulations of a carrier of suitable frequency, 
with the carrier-wave count itself used to identify position along the wires. 
Alternatively, spots on a paper tape could also represent the “bits.” Terms 
assigned to a document would be inserted by a typewriter keyboard posi¬ 
tioning a recording head to the appropriate channel. Reading out would be 
accomplished with a similar keyboard selection of the index term followed 
by a search for coincident bit signals or pulses along the several channels 
selected for search. Carrier signals stored on the channels might serve to 
monitor the reading speed and, via controlled delays, to compensate for 
minor phase shifts to make extremely precise synchronism of the channels 
unnecessary. The result of the search might be preserved in an intermediate 
storage whence it would be used to actuate devices to deliver full-size 
copies, abstracts, or mixtures of the two as instructed by the searcher, or a 
sample of the results could be displayed to guide adjustment of the search 
terms before documents disclosed by the search were delivered. 

Conclusion 

In this chapter, the mechanics of an information search system have been 
described in some detail. Further details, including drawings for construct¬ 
ing some of the devices, can be made available upon request to the authors. 
A number of expressions of interest in the commercial manufacture of equip¬ 
ment for peek-a-boo searching have been received. Most of the devices 
concerned do not appear to be patentable and any patents which do result 
from the work described here will be assigned to the U. S. Government 
and will therefore be available on a royalty-free-license basis. The availa¬ 
bility in this country of simple inexpensive equipment would provide an 
important stimulus to application of Peek-a-boo in fields for which it ap¬ 
pears suitable. 

Extension to Other Fields. The Peek-a-boo principle is versatile enough 
to make it suitable for very small collections as well as for moderately large 
ones. The simplicity of material and equipment enables one to establish a 
file with minimum investment. A variety of inexpensive cards in many 
sizes can be adapted to Peek-a-boo in combination with simple hand 
punches. This simplicity of equipment, together with the great rapidity of 
searching, suggests usefulness in such fields as police files, medical diag¬ 
nosis, law r , catalog and inventory files, real estate, personnel selection, and 
many others. The open endedness with respect to subject which is charac¬ 
teristic of Peek-a-boo, minimizes the losses resulting from errors in setting 
up the index. There is a need, nonetheless, for promulgation of basic index 
lists in a variety of fields which would enable many users to start Peek-a- 
boo files with a minimum of lost motion. 



Chapter 7 

A UNITERM SYSTEM FOR REPORTS 


Charles E. Zenvekh, Jr. 

Research and Development Division, 
Humble Oil & Refining Company, Baytown, Texas 


Introduction 

Reports generated within an industrial concern constitute one of the 
principal sources of information and know-how which can serve as the 
basis for future research and development activities of the organization. 
Since much of the subject content of such reports never gets published, in¬ 
dustrial organizations are confronted by the problem of indexing or classi¬ 
fying their own reports. 

The impetus to undertake application of the Uniterm system to the re¬ 
ports generated by the Research and Development Division of the Humble 
Oil & Refining Company was provided by the rather glaring inadequacies 
of previously used indexing systems. One of these systems was based on 
title indexing. This index was maintained on 3 x 5 inch cards on which, 
together with the index entries, were recorded the usual bibliographic 
data, such as the name of the issuing organization, the date of appearance 
of the report, etc. It was observed—as is so often the case with titles—that 
this form of indexing was often misleading or inappropriate as far as indica¬ 
tion of important aspects of the subject content of reports was concerned. 
Also, the frequent inversion of entries as exemplified by “Cracking, cata¬ 
lytic” was a source of delay and exasperation in using this index. The in¬ 
adequacies of this system led to attempts at more systematic indexing, the 
purpose being to analyze the subject content of the reports more carefully. 
Here again 3x5 inch cards were used for each subject entry set up, and on 
each card were recorded, in addition to the usual bibliographic data, all 
other subject headings under which the report had been indexed. The re¬ 
ports were not indexed in sufficient detail to provide the desired measure 
of accessibility to technical information. At the same time, even with a 
limited amount of indexing—an average of five cards per report—the typ¬ 
ing problem was severe. The number of cards required per report was too 
small to justify mimeographing, but the typing of multiple copies of file 
cards, on the other hand, was time consuming and burdensome. As a con¬ 
sequence there were frequently intolerable delays between the date of 


152 



A UNITERM SYSTEM FOR REPORTS 


153 


issuance of a report and its indexing. Therefore, the Uniterm system 1 was 
adopted to provide faster and more adequate processing of reports. 

Basic Procedures of the Uniterm System 

The operation of the Uniterm system is simple and easily described. A 
separate card is set up for each of the terms used to designate the important 
aspects of reports. On the same card are listed the identifying numbers of 
all reports for which the term in question is an appropriate subject desig¬ 
nation. The report numbers on each card are arranged in the following 
manner for convenience in conducting subsequent searching operations: 
The card is divided into zones as shown in Figure 7-1. The numbers of 
various reports are then arranged in these zones, depending upon the final 
digit in the number. Furthermore, within each of the columns, as Figure 
7-2 shows, the report numbers are arranged in ascending numerical order. 
Before describing the use of a Uniterm file, let us consider how it is estab¬ 
lished. 

The analysis of the subject content of reports by the Uniterm system 
consists of the following steps. The serially numbered report is first read 
and a decision is made as to which terms are appropriate to designate as 
important aspects of its subject content. This review of the report, the deci¬ 
sions as to important aspects of subject content, and their expression by 
appropriate symbolisms—in our case by words—has much in common with 
the performance of alphabetized indexing. The next step is to locate in the 
file those Uniterm cards which correspond with the words that have been 
selected for designating important aspects of subject content. If by chance 
a Uniterm card has not been provided for any one of the words so selected, 
a decision must be made as to one of two possible courses of procedure. 


Absorber; Absorbent; Absorption 


0 

1 

2 

3 

4 

5 

6 

7 

8 

9 

40 

1 

2 

3 

4974 

4695 

146 

1157 

708 

1239 

5150 

41 

262 

123 

5024 


2306 

3737 

868 

1459 


251 

602 

3783 

5064 


2376 

4107 

1548 

1549 


4971 

1042 

4363 

6764 


4796 


4908 

3719 



1542 

4913 

6774 

J 

4906 


5338 

6509 



4262 

5413 



4916 






5412 

5503 



5206 





Figure 7-1. Uniterm card divided into zones (numbered 0-9) in which are entered 
serial numbers of documents which contain the key words or ideas as represented 
by the card heading: “absorber; absorbent; absorption.” The terminal digits of the 
document numbers determine the zone in which the number is entered. 


1 The “Uniterm system” of coordinate indexing was developed by Dr. Mortimer 
Taube and his staff at Documentation, Inc., Washington, D. C. 





154 


PUNCHED CARDS 


Absorber; Absorbent; Absorption 


0 

1 

2 

3 

4 

5 

6 

7 

8 

9 

40 

1 

2 

3 

4974 

4695 

146 

1157 

708 

1239 

5150 

41 

262 

123 

5024 


2306 

3737 

868 

1459 


251 

602 

3783 

5064 


2376 

4107 

1548 

1549 


4971 

1042 

4363 

6764 


4796 


4908 

3719 



1542 

4913 

6774 


4906 


5338 

6509 



4262 

5413 



4916 






5412 

5503 



5206 





Spectra 


0 

l 

2 

3 

4 

5 

6 

7 

8 

i 

9 

120 

1951 

1432 

5713 

1064 

2885 

4576 

1277 

1278 

1279 

190 

2451 

2567 

6263 

2894 

6025 

6026 

2877 

5728 

4489 


2471 

4502 


4524 


6106 

4107 


5069 


6121 

5662 


5714 


6506 

4697 




6421 

6262 




6636 

5707 










6247 







1 



6457 










6517 




Infrared 


0 

1 

2 

3 

4 

5 

6 

7 

8 

9 

80 

71 

72 

443 

1064 

1065 

66 

77 

108 

1249 

1250 

241 

112 

4173 

2894 

2885 

1896 

1897 

118 

1459 

1650 

4091 

1432 

5413 

4144 

4165 

4086 

2017 

4088 

3079 

5530 

4701 

3132 

5723 

5344 

4725 

5266 

2487 

4138 

3539 

6480 

5661 

4652 

6273 

5724 

5295 

5296 

4107 

4388 

4119 



4712 


6764 

5375 


4217 

5278 

4139 



4942 





5267 


4199 



5322 





5297 


4489 



5662 





5537 


5359 



6452 





6457 


5719 










6389 










6449 


! 








6599 


Figure 7-2. A search is conducted for all documents in the Uniterm file which 
contained the key words or ideas: “infrared,” “spectra,” and “absorption. 0 The 
three cards—as shown—are selected, and are compared to determine which docu¬ 
ment numbers are in common to all three cards: viz , number 4107. 




A UNITERM SYSTEM FOR REPORTS 


155 


It may be decided that, for the term in question, a synonym or near-syn¬ 
onym for which a Uniterm card has already been provided may serve to 
designate the subject content aspect. In this case the report number will be 
entered on the existing Uniterm card selected as being appropriate. Alter¬ 
natively, it may be decided to establish a new Uniterm card. This question 
when it arises is an important one, and will be discussed further. After the 
Uniterm cards have been located in the file or new cards provided if neces¬ 
sary, the report number in question is entered on the various cards following 
the procedures described and illustrated in Figures 7-1 and 7-2. 

The Uniterm file is used for searching and retrieving as follows: If it is 
desired to identify all the reports relating to a given subject such reports 
will be listed on a single Uniterm card, provided, of course, the subject in 
question has been set up as one of the Uniterms. In our system, for example, 
one will find listed on a single card all reports embraced by the system which 
relate to the subject “Fungicide.” Many research requirements cannot be 
so simply expressed, however. If a search were directed to a broader sub¬ 
ject, such as biologically active materials, it would be necessary to call 
attention to the entries on two or more Uniterm cards. Thus in this system 
we might consult reports whose numbers had been entered on the Uniterm 
cards for “fungicide,” “insecticide,” “toxicity,” and possibly various other 
cards. 

Another possibility is that we may be interested in fungicide activity, 
only if it involves some one organism or class of organisms. In our file, which 
is devoted principally to petroleum processing, we have not found it ad¬ 
vantageous to index in detail with regard to the organism to w'hich fungi¬ 
cides are applied. The number of reports in the file, however, dealing with 
fungicides is not large. As a result, a person desiring such information is 
well served by the Uniterm file if his attention is directed to this small 
number of reports. If in the future a large number of reports dealing with 
fungicides were to enter our file, it w r ould not be difficult to reanalyze these 
reports, and to set up additional Uniterm cards to correspond to major 
groups of fungus organisms, or individual species of organisms in case they 
were important. This is an important feature of the Uniterm system, and 
the possibility of carrying through a reanalysis of certain reports, at the 
time that such reanalyze becomes necessary, should be kept in mind as an 
important advantage of the Uniterm system. 

The subject contents of most scientific and technical papers usually relate 
to a multiplicity of things. Similarly requirements for information may often 
be stated as involving two or more terms. Thus an information requirement 
might be exemplified as relating to the use of furane chemicals for fungi¬ 
cides. In our system the key terms would be “furane” and “fungicide,” 
both of which appear on Uniterm cards. A list of report numbers which 
might be of pertinent interest to such an information requirement may be 



156 


PUNCHED CARDS 


readily compiled by using this file. This would be done by consulting the 
cards for the terms “Furane,” and “Fungicides,” and noting the report 
numbers that have been listed under both of these terms. There is, of 
course, the possibility that one or more reports may have mentioned a 
furane compound and also referred to fungicides, without the furane com¬ 
pound necessarily having been used as a fungicide. Since our system em¬ 
braces a relatively small number (about 7,500) of reports at the present time, 
the possibilities of being directed to such reports have not been proved to 
be a practical problem. 

At the Humble Oil & Refining Company the reports themselves are main¬ 
tained in a classified arrangement and the reports required for a given prob¬ 
lem or situation may be located by this system. For reports dealing with a 
specific subject, it is often more convenient to utilize the Uniterm card 
file. We maintain an auxiliary card file, arranged numerically by serial 
number of the reports involved. On each of the serial number cards is re¬ 
corded the usual bibliographic data concerning the report in question, as 
well as the class designation which is used for filing the individual reports. 
Thus a report can be located by first identifying it by its serial number in 
the Uniterm files and then using this number to locate its position in the 
classification scheme, and finally locating the report in the classified file. 

Observations Made in Applying the Uniterm System 

Application of the Uniterm system at the Humble Oil & Refining Com¬ 
pany was initiated in the Fall of 1953. It was immediately observed that 
the reports could be indexed much more rapidly by applying Uniterm pro¬ 
cedures than was possible with the subject heading list previously used. A 
three-month backlog of reports awaiting indexing was cleared up in about 
two weeks. Furthermore, it was observed that the Uniterm file very mark¬ 
edly accelerated the speed of searching and retrieving material in the file. 
A quantitative estimate of the time saved is difficult to make, since the 
depth of indexing was markedly increased when the Uniterm system was 
installed. As a result it is now possible to locate material that could never 
have been found in the system used prior to Uniterm. In a brief comparative 
study, searches of three widely different types were completed in one-half 
to one-tenth the time required before installing the Uniterm system. At 
first scientific and technical personnel were a little bewildered by the sys¬ 
tem, particularly by the fact that the customary bibliographic data was 
lacking on the Uniterm cards, and that instead the reports were identified 
exclusively by serial number. A minor amount of explanation, however, 
sufficed to clear up this difficulty, and it was observed that with a little 
experience the technical and scientific personnel were able to use the file 
expeditiously and advantageously. The advantages observed in practical 
application and the inherent simplicity of the system have led to the con¬ 
clusion that its capabilities answer our present report indexing problem. 



A UNITERM SYSTEM FOR REPORTS 


157 


On the basis of experience, we are convinced that the system will continue 
to function well for at least several years to come. When more than 20,000 
reports have been incorporated into the file, it is anticipated that certain 
difficulties may appear. For example, as more and more reports are em¬ 
braced by the Uniterm file, an increasing number of reports will have to be 
listed on the various Uniterm cards. For some of the Uniterms, more than 
one card will be needed to list all the reports. There will be an increase in 
the amount of time required to compare lists of reports pertaining to two 
(or more) Uniterms in searching for reports pertaining both to “Furane” 
and to “Fungicides.” Although such difficulties are foreseeable, it seems 
unlikely that they will seriously impair the usefulness of the Uniterm file 
for several years. At the present time, our Uniterm file embraces about 
7,500 reports, and it is expanding at the rate of about 500 reports per year. 
In addition we are indexing into the file reports written and filed before the 
Uniterm system was started. There are about 5,000 of these reports now 
awaiting our attention. A considerable number of older reports have al¬ 
ready been incorporated in the file. 

It might also be pointed out that our file is restricted almost entirely to 
reports originating with the Humble organization or its immediate affiliates. 
Also included is a small number of confidential reports originating with con¬ 
tractors and with similar persons. Progress reports are not indexed in 
Humble’s Uniterm system. 

As with any system for processing the subject documents for subsequent 
retrieval, a number of policy decisions had to be made. One of these de¬ 
cisions, as mentioned above, concerned the range of material to be em¬ 
braced by the system. Other decisions related to the analysis of the subject 
content of the reports, particularly what aspects of subject content should 
be taken into account and what Uniterms should be used to designate those 
aspects that are considered important. It should be emphasized that it is 
essential to distinguish between these two operations. One of them is the 
decision as to which aspects of subject content are indeed important. In 
this connection, quite naturally, we are guided by the range of interests of 
the Humble Oil & Refining Company. In general, it may be said that the 
reports which originate in our own research laboratories are indexed in 
considerable detail while many reports from affiliates are indexed relatively 
sparsely. For example, at the present time our Company’s limited interest 
in grease and asphalt research would result in reports on these subjects 
being indexed in relatively little detail. Certainly, however, such reports 
would be entered on the Uniterm cards for “Grease” and for “Asphalt.” 
If, at some future time, it should prove advisable to index reports on these 
subjects in more detail, these Uniterm cards would quickly direct our at¬ 
tention to the reports requiring further processing. 

As noted above, once a decision has been made as to what aspects of sub- 



158 


PUNCHED CARDS 


ject content are of importance, there still remains the important decision 
as to selection of suitable terms to designate such aspects. At first individual 
words were used almost exclusively as Uniterms. Thus the number of a 
report discussing methyl alcohol was entered on one Uniterm card for 
“Methyl” and on another card for “Alcohol.” In practice, this procedure 
was found to introduce considerable inefficiency into the operation. A large 
number of entries weie found to accumulate rapidly on the “Methyl” card 
and also on the “Alcohol” card. Furthermore, the need to identify reports 
relating to methyl alcohol occurred sufficiently often that an excessive 
amount of time was being devoted to comparing report numbers entered 
on these two cards. As a result, it was decided to make out a card for 
“Methyl Alcohol.” In this instance, therefore, the term “Methyl Alcohol” 
is one of our Uniterms. It might be pointed out that redoing the file to es¬ 
tablish the card on “Methyl Alcohol” was not difficult. All that was re¬ 
quired was to determine which report numbers appeared on both the Uni¬ 
term cards for “Methyl” and for “Alcohol” and to list these report numbers 
on the card set up for “Methyl Alcohol.” 

It was observed, however, that it was advisable to maintain a generic 
card for “Alcohol.” It was decided further that in the future, we would not 
enter on this card reports that are concerned with “Methyl Alcohol” only. 
Rather a cross reference was provided for the “Methyl Alcohol” and “Al¬ 
cohol” entries in the Uniterm subject list. In this list the “Methyl Alcohol” 
entry is accompanied by the notation, “Methyl Alcohol,” see also “Alco¬ 
hol,” and a corresponding entry under “Alcohol” reads “Alcohol,” see 
also “Methyl Alcohol.” Furthermore, on the “Methyl” Uniterm card we 
note only those reports in whose subject matter the methyl group itself is 
an important entity, as exemplified by the methyl radical, CH 3 -. Enough 
has been said, perhaps, to make the point that the Uniterm system must be 
applied with care and understanding to achieve optimum results. 

A similar situation was observed in connection with the Uniterm cards 
originally established for “Catalytic” and for “Cracking.” It quickly be¬ 
came apparent that it would be advantageous to set up a Uniterm card for 
“Catalytic cracking” for much the same reasons that a card was set up for 
“Methyl alcohol.” However, just as we found it advantageous to maintain 
an “Alcohol” Uniterm card but to restrict its use to alcohols other than 
methyl alcohol, we found also it advantageous to maintain a “Cracking” 
Uniterm card, but to restrict its use to instances not covered by “Catalytic 
cracking.” Corresponding cross-references for “Catalytic cracking” and 
“Cracking” were set up on our standardized list of Uniterms along the 
lines outlined above for “Methyl alcohol” and “Alcohol.” Our technical 
men have pointed out a number of similar instances in which it is advisable 
to provide a Uniterm card for such terms as, for example, “Solvent extrac¬ 
tion.” In this instance, however, the number of all reports that pertain to 



A UNITERM SYSTEM FOR REPORTS 


159 


Bromination , see also halogenation 

Bromine , see also halogen 

C/H, see carbon plus hydrogen plus ratio 

Chlorination, see also halogenation; sulfochlorination 

Chlorine, see also halogen 

Concentrate; concentration; concentrator 

Condensate; condensation 

Condensed; condenser 

Digital; digitization; digitizer 

Gage; gaging 

Gauge , see gage 

Halogen , see also bromine; chlorine; fluorine; iodine 
Halogenation, see also bromination; chlorination; sulfochlorination 
Liquid; liquor 

Monobromoxylene, see halogen plus xylene 

Monomer, see also monoolefin; specific monoolefin such as ethylene, propylene, butyl¬ 
ene 

Paraxylene, see p-xylene 
PCP, see pentachloraphenol 
Toxicity; toxicology 
Versene, see chelate 

Figure 7-3. Some representative subject headings and cross-referencing proce¬ 
dures. 

"solvent extraction as well as to other types of extraction are being entered 
on the “Extraction” card. In addition, we are entering the appropriate 
report numbers on both “Solvent Extraction” and “Solvent” cards. Thus 
the “Solvent Extraction” card may be regarded as providing a ready made 
listing of report numbers that are entered both on the “Solvent” and the 
“Extraction” cards. This further illustrates the way in which we have care¬ 
fully controlled the use of terminology to insure the effectiveness of the 
Uniterm system. 

In applying this system, we have found it highly advisable to exert care¬ 
ful control of synonyms in selecting terms to be set up on Uniterm cards. 
We have drawn up a list of terms for which Uniterm cards have been estab¬ 
lished, and we have found it to be essential in insuring consistency in the 
use of terminology, particularly when consulting the Uniterm file to identify 
reports of pertinent interest to a given problem or situation. 

During the initial phases of applying the Uniterm system, there was a 
marked tendency for different synonyms to be set up on individual Uni¬ 
term cards. Such situations may be exemplified by a portion of our list of 
the terms for which we have Uniterm cards in our files. As shown in Figure 
7-3, such closely synonymous terms as halogen, chlorine, and sulfochlorina¬ 
tion obviously require reconsideration, if use of the file is not to involve 
either (1) excessive time in consulting a number of cards to cover a single 
subject or (2) failure to cover a single subject by overlooking the need to 
check a number of Uniterm cards. For such synonyms and near-synonyms, 



160 


PUNCHED CARDS 


it is necessary, for reliability and efficiency, to provide a single Uniterm 
card and to be careful to see that reports relating to the subject in question 
are entered on the card for the term selected. 

In Figure 7-3, various sets of synonyms and near-synonyms have been 
drawn together into adjacent positions by alphabetization of the various 
words. It must be noted, however, that alphabetizing cannot be relied upon 
to direct attention to all instances in which synonyms or near-synonyms 
should be grouped and a single Uniterm card prepared for the group of 
terms. Thus, for example, we have found it advisable to prepare a single 
Uniterm card for “Doucil” and “Zeolite” and also a single Uniterm card for 
“Drainings” and “Sludge.” In setting up such cards some one of the syn¬ 
onymous or closely-synonymous terms is selected for the Uniterm card, and 
the synonyms and near-synonyms are cross-referenced in the standardized 
listing of terms, as shown in Figure 7-3. The advantages of providing 
“(See also ” references have been discussed previously in connection with 
setting up Uniterm cards for “Methyl alcohol,” “Catalytic cracking,” 
and “Solvent extraction.” It is perhaps obvious that one may learn a great 
deal from analogous procedures as practiced in subject indexing. We have 
found it highly advantageous to set up our Uniterm cards for chemical com¬ 
pounds on the basis of the well-established nomenclature practices of 
“Chemical Abstracts.” Thus the name of a compound which consists of 
two or more words, e.g., methyl alcohol, is regarded as a single Uniterm 
and such terms are not split into component words when establishing Uni¬ 
term cards. Occasionally, we find it helpful to our scientific and technical 
personnel to take into account the abbreviations and chemical slang that 
they use habitually. Thus we may use “MPK” as a term to denote methyl- 
propyl ketone. 

At the present time, the terms which are included in our lists and for 
which Uniterm cards have been prepared number about 3000. It seems likely 
that this list will increase rather slowly in the future, with most of the fore¬ 
seeable expansion being devoted to cross-reference entries of the “see” or 
“see also” types. 

Conclusion 

Our experience with the Uniterm system might be summarized as fol¬ 
lows: We have found that it simplifies and expedites our indexing operations- 
Furthermore, accessibility to the information in our reports has been con¬ 
siderably improved. However, in applying the Uniterm system, it is highly 
advantageous, if indeed not necessary, to insure that terminology is used 
in a consistent fashion. Careless use of terminology may result in reports of 
pertinent interest to a given subject, problem, or project being scattered 
over a range of Uniterm cards, with the possibility that such material 
may be overlooked when making a study. 



Chapter 8 

AN IMPROVED ANESTHESIA 
RECORD CARD* 

Max S. Sadove, M.D. and Myron J. Levin, M.D. 

University of Illinois, Chicago, III. 

AND 

Veterans Administration Hospital, Hines, III. 

The search for a simple, satisfactory flexible system of maintaining an¬ 
esthesia records has been going on for at least a decade. Each newly devised 
system has been an improvement over the previous ones, only to become 
itself obsolescent as new anesthetic agents and techniques are introduced. 
Certain basic requirements of the joint council on Hospital Accreditation 
and other official bodies must be met. A complete anesthesia record must 
be filed within the patient’s clinical chart. The clinical records should be 
able to serve as aids in teaching and learning and they should also be valid 
as medico-legal documents. Each record must be specific for the job it per¬ 
forms; but in addition, it must permit future expansion as the particular 
field it serves expands. 

Anesthesia records should be so simple that a competent secretary or a 
medical student can complete them and understand them. This is especially 
necessary because even the trained anesthesiologist cannot give his fullest 
attention to his patient while completing a complicated record. 

Another essential is that information be recorded in such a manner that 
it will be available for future use in compiling statistics as well as in study¬ 
ing a given case. 

Finally, in a given system of anesthesia records, it is important that space 
be available for all current technics, agents, methods and all data about 
the patient such as his physical condition, pre- or postoperative complicat¬ 
ing factors, cause of death, etc. It is important that provision be made in 
the chart for everything that may possibly occur during the conduct of a 
case. In the previously existing anesthesia records the gathering of statistical 
data regarding preoperative and postoperative complications has been par¬ 
ticularly difficult. Newly developed technics such as the use of hypo- 

• Acknowledgment is due to Mr. Erwin L. Vaughn of E-Z Sort Systems, Ltd., for his 
major part in the development of the Illinois Anesthesia Punch Card. 


161 



162 


PUNCHED CARDS 


thermia, hypotension or mechanical respiration-assisting or controlling 
devices which have come to be accepted standard anesthetic technics, 
cannot be easily coded on the preexisting forms. 

With the above criteria in mind, a new anesthesia record form was de¬ 
vised, which we believe fulfills most of the requirements. 

An edge punched card could, if properly designed, fulfill all of the cri¬ 
teria for anesthesia records listed above. Several such cards have been used 
successfully for the past few years in various institutions throughout the 
country. The Chicago Keysort card was taken as a basis from which the 
present card was developed. At a casual glance this card appears exceed¬ 
ingly complicated. Yet on closer inspection, the seeming complexity re¬ 
solves into its most important advantage. This is the fact that a minimum 
of writing is necessary to complete the required entries. Almost every con¬ 
ceivable presently used technic, agent, and complication is printed on the 
card, requiring only a minimal amount of checking. Once the user is familiar 
with the card it becomes very simple to maintain records with it. The card 
is so constructed that the front and back sides are each separate units, 
making it unnecessary to turn the card during the conduct of an anesthetic 
procedure. The front will be discussed first and then the reverse side. 

The front or face of the card contains the following data (see Figure 8-1): 

Name 

Physical Status 

Age 

Sex 

Ward 

Room 

Register Number 
Anesthetic Number 
Date 

Proposed Operation 
Preoperative Diagnosis 
Pre-Anesthetic Medication 
Identity 

Positive Findings 

Preparatory Therapy 

Graphic chart with code for Graphing 

Remarks column for items not coded or listed 

Agents: Primary and Secondary for Induction, Maintenance and Emergence 
Methods: Induction, Maintenance and Emergence 
Types of Airway Maintenance 
Surgeons 
Anesthetists 
Postoperative Diagnosis 
Operations Performed 
Left-hand margin (punched) 

Sex 

Special Interest 



AN IMPROVED ANESTHESIA RECORD CARD 


163 


Anesthesia Time 
Operation Time 
Special Studies 
Anesthetist 
Preparatory Therapy 
Physical Status 
Age of Patient 
Year 

Right-hand margin (punched) 

Site of Operation 
Position of Patient 
Level or Plane 
Classification of Operation 
Death Analysis 
Sections “B” <fc “C” 

The front of the new anesthesia record card is 10^ inches in width and 8 
inches in height. The central portion of the card is a graphic chart similar 
to those used by most anesthesiologists, but with several important ad¬ 
vantages over most of them. These are as follows: 

1. The graph extends over a 4^-hour period whereas most of the pre¬ 
vious cards do not exceed 2^2 to 3 hours. This permits the use of a single 
card for most operations, whereas previous cards quite frequently required 
a second sheet. If the operation is shorter than 2 hours nothing is lost since 



Figure 8-1. Illinois E-Z sort anesthesia punched card (front). 































164 


PUNCHED CARDS 


a record is required even for a 10 minute procedure. The increasing fre¬ 
quency of lengthy operative procedures makes this feature of the card finan¬ 
cially attractive as well as convenient for the user. It also conserves filing 
space. 

2. Six blank lines are provided for the designation of Agents. This was 
done to save space, since all the agents could not be listed by name. It offers 
the additional advantage in that the primary or most important anesthetic 
agents can be listed first with supplemental agents given on the lines below 
or, if desired, these may be listed chronologically with the first one used 
on the top line and each succeeding one on the fine below. Even in these 
days of polypharmacy it is rare that more than four or five agents would 
be employed in the conduct of any single anesthetic, so that the use of six 
lines leaves ample room for even the most extreme cases. 

3. Two separate lines are provided for the graphic representation of 
muscle relaxant drugs since these are now extremely important in the con¬ 
duct of anesthesia, but are considered to be adjunct drugs rather than 
anesthetic agents per se. A box at the end of each fine is provided for re¬ 
cording the total dose of these drugs. 

4. Two lines are available for recording intravenous therapy—saline, 
dextrose, blood, etc. These are to be used in the manner described by other 
authors, but reiterated here: 

-represents blood volume expanders and plasma 

'—~ represents watery solutions such as dextrose or physiologic saline. 

1 represents whole blood. 

In each instance, the addition of a new bottle of solution is marked with a 
vertical arrow, thus ‘ f * and the kind and quantity of solution is written 
18 ga * 5% D/W NSS 

in. For example: RA | 1000 } 500 would indicate that 1000 

ml of 5% dextrose in water was started via an 18-ga. needle in the right 
arm at the first vertical arrow, and that 500 ml. of physiologic saline solu¬ 
tion was added via the same infusion set at the second arrow. Blood or other 
types of fluids can be graphed in a similar fashion as given. Since two in¬ 
fusions in separate veins are not infrequently employed two lines are left 
for this purpose. 

5. The next section represents analgesia, beneath which are four lines 
for graphic representation of plane or level of surgical anesthesia. The 
anesthetic level may be followed easily during the conduct of a given case 
if it is thus graphically recorded. These sections of the chart apply equally 
well to regional as to general anesthesia technics. Since definite stages and 
planes of general anesthesia are defined, the line on the graph goes lower as 
the anesthetic level becomes deeper. In regional or spinal anesthesia, the 
level of the body segment reached by the anesthetic at a given time can 



AN IMPROVED ANESTHESIA RECORD CARD 


165 


similarly be recorded. The levels usually represented in spinal analgesia 
are: (1) below the twelfth thoracic segment; (2) below the seventh thoracic; 
(3) below the fourth thoracic or nipples; (4) above the fourth thoracic. 

6. Immediately below this is the blood pressure, pulse, and respiration 
graph. To the left, is a box illustrating some of the more commonly used 
symbols. These are the symbols most commonly used by anesthesiologists 
to indicate various happenings graphically. It is a specialistic shorthand 
which is widely understood among members of the specialty. These are in¬ 
cluded to make interpretation easier to the uninitiated. It is not an all-in¬ 
clusive list. When other symbols are used, they may be explained in the 
“remarks” column to the right of the graph. 

The space allotted to “Remarks” is smaller than that ordinarily found 
on most record forms. This was made possible by the specific inclusion in 
the printed portion of a large part of the material which is frequently en¬ 
tered under “Remarks.” 

7. The bottom line of the graphic section indicates whether or not CO» 

absorption was employed. When “total” absorption is employed, i.e., with 
the closed system and soda-lime canister, the boxes in this line are com¬ 
pletely cross hatched thus: When a semi-closed system with par¬ 

tial absorption is used, the boxes are partially filled, thus: BBB. In a 
completely open system, they are left empty. 

8. Immediately beneath the graph is a line for indication, by means of a 
vertical arrow, the time of specific remarks, thus —“ j \ X 2 T 3 ”- These are 
then entered in the “Remarks” column to the right of the graph, with the 
same key number being employed, e.g., “3—Neo-Synephrine 0.5 mg I.V.” 

9. The next two lines permit entry of induction, maintenance and emer¬ 
gence anesthetic agents with sufficient space for at least six different agents 
(important in these days of polypharmacy or balanced anesthesia). Be¬ 
low this, lines are provided for indication of anesthetic methods, means of 
airway maintenance, etc., and the usual identification of surgeons, anesthe¬ 
tists, postoperative diagnosis and operation performed. 

The face of the card has 2 rows of punch holes along the right and left 
margins. The use of these will now be explained, beginning with the upper 
left-hand comer of the card. It should be emphasized at this point that all 
the information to be punched is contained in the central portion of the 
card. Thus, punching does not have to be done during the operative pro¬ 
cedure, but can be done later either by the anesthetist himself or by a clerk 
or other person designated to do this work. 

The holes along the left margin are punched as follows: 

1. Female. The deeper hole is to be punched for all female patients. No 
punch is necessary for male patients. 

2. Spec. Int. This section will be punched either deep or shallow, at the 



166 


PUNCHED CARDS 


direction of the particular department head, to indicate cases of unusual 
interest. It is an identifying means for quickly sorting the unusual from the 
more routine cases. These holes may be used for different purposes by dif¬ 
ferent groups and at various times by the same group. For example, cases 
of unusual surgical interest such as rare tumors, extraordinary surgical pro¬ 
cedures, etc., may be punched shallow, while cases of particular anesthesia 
interest may be punched deep. 

3. Anesthesia Time. This section is used to indicate the time in hours 
that the patient was anesthetized. This, and the next section, operative 
time, are punched by numerals, as indicated—for less than x /i hour, a 
shallow punch is made in the top space; for less than 1 hour, a deep punch 
is made. These sections provide a more extensive breakdown than some of 
the older records, in that 1, Ij-i hour cases can all be indicated. Cases 
longer than 5 hours are quite infrequent, and it is relatively unimportant 
to distinguish between a 6 and a 9 hour case because both are extremely 
prolonged surgical procedures and represent less than 3 per cent of all 
surgical cases. 

4. Operative Time. Same as above, for anesthesia time. 

5. Special Studies. This section is distinguished from section 2 above in 
that it is used for recording particular research studies rather than just 
interesting or unusual cases. Each study may be assigned a specific number 
which is punched as directed by the head of the department. As many as 
twenty different research studies can be carried on simultaneously and by 
means of reassignment of given numbers after a significant lapse of time 
the section can be used almost indefinitely. This section can also be useful 
in re-identifying cases to be studied after the card has been punched. For 
example, if it were desirable to study all cases of female patients below the 
age of 35 who had appendectomies performed with ether, this would ordi¬ 
narily require sorting in four sections of the chart—sex, age, site of opera¬ 
tion and anesthetic agent—whereas if such cases were arbitrarily assigned 
a number in the special studies section they could be found in a single sort. 
Other examples of special studies might be “cardiac arrest” cases, the “ether 
analgesia” technic, or hypnosis, which is again undergoing a period of in¬ 
creasing popularity. Each of these could be assigned a specific number in 
this section of the chart and the records could thus be made readily avail¬ 
able. This section is coded as shown in Figure 8-2. 

6 . Anesthetist. The responsible anesthetist in each case will punch his 
own code number here. There is no necessity to learn more than this one 
code punch for each person. In the case of residents, the numbers can be 
reassigned when the resident leaves and the dates will distinguish one from 
another. This section is likewise coded as shown in (Figure 8-2). 

7. Preparatory Therapy. This section is used for recording maintenance 



AN IMPROVED ANESTHESIA RECORD CARD 


167 


0000|| 000.0. OOOMI 00.00. oo.o.l 

ooooVV oooloV oooool o 6 9 o o I oooooV 

10 7 4 2 1 0 10 7 4 2 1 0 10 7 4 2 1 o 10 7 A 2 1 o 10 7 4 2 1 0 

( 1 ) ( 2 ) ( 5 ) ( 4 ) ( 5 ) 

0 ° • 0 I ° I 0 ° °| o.oo.| O.o.o. loooofe 

oooooV oVooof oooooV oooool Vooool 

10 74210 10 74210 10 74210 10 74210 10 74210 

( 6 ) ( 7 ) ( 8 ) ( 9 ) ( 10 ) 

looolo looloo looooo Volooo looooo 

10 74210 10 7 4210 10 74210 10 7 4210 10 7 4 210 

( 11 ) ( 12 ) ( 15 ) ( 14 ) ( 13 ) 

.0..00 0 . 0 0 0 0 1 . 0 0 . 0 ..O.O. IMOOO 

looooo lloooo looooo looooo looooo 

10 74210 10 74210 10 74210 10 74210 10 74210 

( 16 ) ( 17 ) ( 18 ). ( 19 ) ( 20 ) 

Figure 8.2. Explanation of coding Illinois E-Z Sort Systems anesthesia card. Spe¬ 
cial studies and anesthetist: The same code applies to both of these fields. These fields 
permit the direct extraction of any numeral from 1 to 20 inclusive, coded as shown 
above. Two sorting needles are required to extract numerals 1, 2, 4, 7, 10, 11, 12, 14, 
17. Three needles are required to extract numerals 3, 5, 6, 8, 9, 13, 15, 16, 18, 19, 20. 


therapy with drags which may affect the conduct or outcome of anesthesia 
or surgery. This is distinguished from preanesthetic medication in that it 
represents maintenance therapy such as cortisone, digitalis, insulin, etc., 
which was employed preoperatively. This section is an innovation in the 
present card. To our knowledge it has not been employed previously and 
therefore its use may require somewhat more detailed explanation than 
other parts of the card. It was felt necessary to include such a section be¬ 
cause the advent of a large number of newer drugs in recent years has com¬ 
plicated the administration of anesthesia. Many of these drags such as the 
newer hormonal preparations can cause serious physiologic disturbances 
when anesthesia is subsequently administered, unless the anesthesiologist 
has been informed of their use and has taken proper precautions. It is not 
the purpose of this treatise to discuss individually the various therapeutic 
agents which may effect the conduct of anesthesia. Five of these have been 
listed on the card, and space has been left for seven others which may be 
added to the armamentarium at some future time. The five currently listed 
include digitalis, the narcotic drags, antihypertensive and tranqulizing 
drugs, barbiturates, and endocrines. Other drugs such as radioactive iso¬ 
topes, nitrogen mustards and similar antimalignancy agents, anabolic 
agents, and other types of chemotherapy which may alter the conduct of 



168 


PUNCHED CARDS 


0 0 0 | 

O O | 0 

0 0## 

0 | 0 0 

O • 0 • 

0 0 0 1 

0 O 1 0 

0 0 o o 

0 9 0 0 

0 0 0 0 

7 4 2 1 

7 4 2 1 

7 4 2 1 

7 4 2 1 

7 4 2 1 

(1) 

(2) 

(5) 

(4) 

(5) 

0 • • 0 

IOOO 

• o o • 

• 0 • 0 

• •00 

e o o o 

1 0 0 o 

0 0 0 0 

O 0 O 0 

0 0 0 0 

7 4 2 1 

7 4 2 1 

7 4 2 1 

7 4 2 1 

7 4 2 1 

( 6 ) 

(7) 

(6) 

(9) 

(0) 


Figure 8-3. Explanation of coding. Year and physical status: The same code ap¬ 
plies to both fields. The above field permits direct extraction of any numeral from 
1 to 9 coded as shown above. A single sorting needle will extract numerals 1, 2, 4, 7. 
Two sorting needles are required to extract numerals 3, 5, 6, 8, 9, 0. 

anesthesia may be assigned punches in the blank spaces provided in this 
section. 

8 . Physical Status. This is the standard A.S.A. physical status groups I 
thru VII. These are coded as shown in Figure 8-3. 

Official American Society of Anesthesiologists’ Classification 

Non-emergency: 

Class I. Patient has no systemic disease (though he may have a simple 
hernia or other nonsystemic disorder). 

Class II. Patient has a mild systemic disease (such as moderate anemia, 
history of heart disease without symptoms or signs, mild diabetes under 
good control, etc.). 

Class III. Patient has a moderately severe systemic disease which is not 
yet a threat to his life (such as heart disease with moderate symptoms, 
early carcinoma, severe diabetes under control, or such combinations as 
anemia plus mild diabetes, or chronic bronchitis plus dehydration, etc.). 

Class IV. Patient has a severe systemic disease which is already a threat 
to his life (such as heart disease with failure, uncontrolled diabetes, or com¬ 
binations of disorders that are a threat to his life, etc.). 

Emergency: 

Class V. Such emergency cases that would otherwise be classed in I or II. 
Class VI. Such emergency cases that would otherwise be classed in 
III or IV. 

Class VII. Moribund patients (such as the accident case that is in irre¬ 
versible shock, severe bowel obstruction of many days duration, etc.) 
This category will rarely be used. 

Mark the appropriate number in the Physical Status Space. In a good 
many anesthesia records the term “risk” or “operative risk” is used and a 



AN IMPROVED ANESTHESIA RECORD CARD 


169 


designation is given for evaluation of this. In evaluating operative risk a 
good many factors must be considered, of which physical status is but one. 
In addition to the physical status of the patient, the kind and duration of 
the operative procedure must be taken into account. A patient who might 
be a normal risk for removal of an ingrown toenail under local anesthesia 
could conceivably be a very poor risk for major surgery such as a pneumo¬ 
nectomy. 

The skill and experience of the surgeon should also be considered in 
evaluating the operative risk. The more skillful the surgeon is in perform¬ 
ing the proposed surgical procedure, the less the operative risk. This also 
applies to the anesthesiologist. One who is versatile and well trained can 
carry a patient through much more difficult surgery with less risk than can 
the person of less experience. 

Finally, duration of anesthesia and operation are important factors in 
risk. The surgeon who can perform a given procedure in one hour will en¬ 
counter noticeably less morbidity than will one who requires four hours for 
the same operation. This then becomes a factor in risk. 

It is because of the innumerable variables associated with “risk” and the 
changes that can occur in these from time to time even during the conduct 
of an anesthetic that the authors decided it was of little value from a sta¬ 
tistical viewpoint. We, therefore, chose to use “physical status” which is 
always the same for a given patient on a given day regardless of whether 
or not operation is contemplated. 

9. Age of Patient. This section differs only slightly, but significantly, 

from previous cards. The ages recorded are in years. The first punch (—1) 
covers all infants from newborn to 1 year of age. The second (—5) covers 
all patients from 1 to 5 years. These pediatric cases have largely been 
neglected in previous recording by decades, but from the point of view of 
the anesthesiologist the young child presents very different problems from 
the older child or adult. It was therefore decided to include this group in a 
separate punch. From here on age is classified by decades up to age 70. 
Patients over 70 years old are generally classified as geriatric cases so that 
a further breakdown was not considered important. ' 

10. Year. This is a coded section, but again, only one coding is used by 
all individuals concerned, for an entire year, so that cards can be pre¬ 
punched prior to use if so desired. This is coded in the same manner as 
“physical status” above. 

The right-hand side of the card requires somewhat less explanation for its 
proper utilization: 

1 . The eight upper portions of this side of the card refer to site of opera¬ 
tion. These should be self-explanatory. They are similar to the designations 
employed in the Chicago Keysort and similar edge-punched cards. 



170 


PUNCHED CARDS 


2. Position, again self-explanatory—refers to the position of the patient 
during the operative procedure. Nine of the most common positions are 
indicated with specific punches and a blank is provided for unusual posi¬ 
tions not ordinarily employed. 

3. Level or Plane. A shallow punch refers to plane of surgical general 

anesthesia and deep punch refers to height or level of spinal or regional 
anesthesia. • 

4. The next sections refer to type of operative procedure—diagnostic, 
therapeutic, definitive, palliative, major or minor. Conceivably a given 
operation might be—therapeutic, definitive and major. These have all 
been available for indication on previous record forms but the authors feel 
that the arrangement has been such that they were not utilized to their 
fullest capabilities. Here, they have been placed in a single section of the 
card and can easily be punched. 

5. Anesthesia or Analgesia. This refers to the type of services rendered 
by the anesthesiologist. It will usually be the former, but in such cases 
as bum dressings, painful examinations, first and second stages of labor, 
etc., analgesic technics may be employed. Analgesic technics may also be 
used in therapeutic nerve blocks for pain, etc. 

6 . Death. This section classifies deaths as preventable or nonpreventable, 
and as to whether or not anesthesia contributed. Accurate use of this sec¬ 
tion requires a full discussion of the case by all concerned prior to punching. 
It should not be abused by punching without mature consideration of all 
deaths. If an autopsy has been performed its results should be known to the 
anesthesiologist before this section is punched. If an autopsy has not been 
performed the entire clinical record, including the anesthesia chart, should 
be carefully reviewed and the department should sit in judgment as to 
whether death was preventable or non-preventable and as to whether or 
not anesthesia contributed. It is rare that more than a few deaths would 
occur in a month’s time in any single department. It is logical therefore to 
suggest that all deaths be discussed in a monthly department meeting and 
that the cards be punched after the meeting. 

7. Sections “B” and “C” are unassigned and may be utilized in any 
manner desired by a specific department head. For example “cardiac 
arrest” cases and such things as neurological complications of spinal an¬ 
esthesia, specific drug reactions, cases of surgery performed by distinguished 
surgeons visiting the hospital or any other desired data might be recorded 
here. 

The reverse side of the card (Figure 8-4) is placed in a vertical position 
so that it can also be punched along the right- and left-hand margins. 
Four rows of punch holes are provided along each margin. The upper central 
portion of the card provides a number of blank lines for entering preopera- 



.4.V IMPROVED ANESTHESIA RECORD CARD 


171 


tive and postoperative summaries. Starting in the upper left corner the 
punches on this side of the card are as follows: 

Left Hand Margin (Punched) 

Pre-Anesthetic Medication 
Relaxants A Their Antagonists 
Vasopressors 
I.V. Therapy 

Pre- and Post-operative Complications 
Right-Hand-Margin (Punched) 

General Anesthesia Technics 
Endotracheal Technics 
Spinal Technics 
Regional Technics 
Hibernation, Hypothermia, etc. 

Regional agents 
General Anesthetic Agents 
Anesthesia Complications 

1 . Pre-anesthetic Medication. Spaces are provided for drugs, route of 
administration, and effect. The agents and technics usually employed are 
printed on the card, and 13 blanks have been provided for additional ones. 

2 . Relaxants. Here again the usual ones are listed with 7 blanks provided 
for additional agents. 

3. Vasopressors. This lists the usual drugs plus blank space, with space 
for recording dosage, technic of administration, and reason for its use. 

4. 7.F. Therapy. This section permits recording of blood, plasma volume 
expanders, electrolytes, sugars, etc., with 14 blank spaces plus the more 
commonly used agents. 

5. The next section extends to the bottom of the left-hand side of the 
card. It permits punching of all pre- and post-anesthetic complications 
plus recording of specific complications. An ingenious punching device per¬ 
mits punching either preoperative or postoperative complications, or both, 
plus an additional line for deaths. The first hole represents preoperative 
complications; the second hole represents postoperative complications. If 
both pre- and postoperative are present in the same category, the third hole 
is punched, and if death ensues, the fourth is punched. For example, a 
patient with pulmonary TBC, active, would have the first hole punched 
under Reap. Major. If this same patient then developed atelectasis post- 
operatively the third hole would be punched. Whereas had the patient 
developed atelectasis without the pre-existing TBC the second hole would 
have been punched. 

Preoperative complications may be punched at the time of the preopera¬ 
tive visit to the patient. If the same patient later develops postoperative 
complications this could then be easily recorded by punching the third 



172 


PUNCHED CARDS 



Figure 8-4. Illinois E-Z Sort anesthesia punched card (back). 


instead of the second hole. If death later ensued it could still be indicated 
by punching the fourth hole. The specific complications are not individually 
punched but the more common ones are grouped according to systems and 
printed on the lines opposite the system punched. For example, under the 
heading “NEUR DISEASES” is listed “psychosis.” If a patient developed 
a psychotic episode postoperatively this section would be punched in the 







































































AN IMPROVED ANESTHESIA RECORD CARD 


173 


second hole and psychosis circled with a pencil with the notation “P.O.” 
after it. In the event the patient was a known epileptic and the anesthetic 
resulted in no aggravation of his epilepsy, the first hole would be punched 
and the word EPILEP circled with the notation “Pr.OP” after it. 

Each of the sections in this grouping is handled in a similar manner. The 
specific complications are circled or written in and the applicable punch is 
made for the group under which a specific complication is classified. 

It is true that if one were searching for cases of a specific complicating 
disease, for example—“Nephritis”—it would be necessary to extract all 
cards for “GU diseases” and then hand sort for “Nephritis.” The number of 
cards to be hand sorted would not ordinarily be unwieldly and in the event 
that it were knowrn in advance that a certain specific disease or condition 
was to be studied, the punches “B” or “C” on the face of the card could 
then be utilized for this purpose. 

For example, in a mental hospital it is to be expected that a good pro¬ 
portion of all patients would have neuropsychiatric disorders. In the event 
it was desired to study all epileptics in such a hospital, “B” punched shallow 
on the face of the card (lower right comer) could be assigned to epilepsy 
and this would tend to simplify sorting. Punch “C,” if desired, might like¬ 
wise be assigned to manic depressive psychosis. 

The right-hand border on the reverse of the card represents technics and 
agents of anesthesia plus complications occurring during administration. 

1 . The topmost section includes technics of general anesthesia. The four 
holes at each technic represent, respectively, (1) induction, (2) maintenance, 
(3) emergence, and (4) supplemental. (The latter refers to technics used to 
supplement the maintenance with another agent (for example, intravenous 
supplementing inhalation). 

2. The second section classifies the types of tracheal intubation. 

3. The next two sections include technics of spinal and regional blocks. 

4. Anesthetic agents are listed next. The regional agents are given first, 
with suitable blank spaces for the newer drugs. The agents for general 
anesthetics are given, provided with classifying punches, viz: (1) primary, 
(2) induction, (3) supplementary, (4) emergence. 

5. Finally, a section is provided in the lower-right hand corner for re¬ 
cording anesthetic complications and the time of their occurrence during 
induction, maintenance, emergence or following premedication. 

The reverse of the card can be marked and punched at any time after the 
case has been concluded. We believe this part of the card will furnish the 
most valuable statistical information. 

We also feel that much needless searching of records for statistical data 
can be readily avoided by the use of this card. For hospitals interested in 
less detailed information, the face of the card presents a readily available 
medico-legally complete record and this side alone may be used. 



174 


PUNCHED CARDS 


References 

1. Tovell, R. M., and Dunn, H. L., “Anesthesia Study Records/* Anesthesia & 

Analgesia 11 , 37-41 (Jan.-Feb. 1932). 

2. Rovenstine, E. A., “Method of Combining Anesthetic and Surgical Records for 

Statistical Purposes/* Anesthesia & Analgesia , 13 , (May-June 1934). 

3. Nosworthy, Michael, “A Method of Keeping Anesthetic Records and Assessing 

Results/* Brit . J . Anesth 17, 160-179 (July 1943). 

4. Pender, John W., “A Combined Anesthesia Record and Statistical Card/* Anes¬ 

thesiology , 7, 606-610 (Nov. 1946). 

5. Conroy, W. Allen, Cassels, W. H., and Stodsky, Bernard, “The Chicago Keysort 

Anesthesia Record/* Anesthesiology , 9 , 121-133 (March 1948). 

6. Sadove, M. S., and Levin, M. J.,“The Illinois E-Z Sort Anesthesia Record Card,” 

Anesthesiology , 19 , 178-187 (March-April 1958). 



Chapter 9 


PUNCHED CARDS AS AIDS TO QUALI¬ 
TATIVE CHEMICAL ANALYSIS BY 
SPECTRAL METHODS* 


L. E. Kuentzel 

Wyandotte Chemicals Corporation 
Wyandotte, Michigan 


Introduction 

Qualitative chemical analysis employing physical methods, stripped of 
confusing but necessary trimmings, almost always involves the measuring 
or more or less distinctive physical properties of a compound and comparing 
them with similar physical data obtained from compounds of known purity 
by the same reproducible methods. Many such physical properties are 
simple to measure and have been used for identification purposes for some 
time. Among these are such properties as the boiling or melting point, index 
of refraction, density, crystalline form and optical activity. Moreover, it is 
easy to record and tabulate such data for the preparation of tables of stand¬ 
ard data for identification purposes. However, the physical properties of 
compounds as measured by the spectral methods of absorption spectro¬ 
scopy, x-ray diffraction, mass spectroscopy, raman spectroscopy and others 
are of such complex nature that they cannot be reduced to a simple number. 
In such cases one must compare complex sets of data with standard sets of 
similar data to identify the unknown material being analyzed. Although the 
very complexity of such sets of data provides highly desirable details for 
correlation purposes it does present problems in connection with tabulating, 
sorting, comparing and distributing the data in the normal course of quali¬ 
tative analysis. This has led to an increasing use of notched-type punched 
cards. Also, more recently, under pressure of rapidly accumulating tens of 
thousands of such sets of standard data the use of International Business 
Machine cards and sorting equipment has become extremely helpful. This 
chapter describes the practical application of punched cards, both the 

* Thanks are due the American Society for Testing Materials for permission to 
reproduce in this chapter its copyrighted Codes and charts pertaining to ASTM- 
Wyandotte punched cards for indexing spectral absorption data. Thanks are due the 
Consolidated Electrodynamics Corp., Pasadena, California, for permission to repro¬ 
duce in this chapter its copyrighted mass spectrum card. 



176 


PUNCHED CARDS 


notched or Keysort type and the punched or IBM type, to the problems 
of indexing and correlating complex sets of physical data for qualitative 
analytical purposes and presents the codes and instructions for the most 
widely used systems. 

General 

There is no need in this chapter to discuss at length the fundamentals of 
preparing and using punched cards. Chapters 2 and 3 provide descriptions 
of such operations. Therefore, attention will be paid to particular examples 
of the uses of such cards. Before discussing such applications, a brief com¬ 
parison of the relative advantages and disadvantages of the two types of 
cards in this particular field will assist the reader in evaluating the most 
desirable method for his specific problems. 

Hand-Sorted Cards. The chief advantage of the hand-sorted card lies 
in the fact that standard data for which one is searching may be printed or 
written on it so that once the proper card is located much of the pertinent 
information may be read directly from it. Also, sorting operations can be 
carried out by hand with uncomplicated and inexpensive equipment. The 
notching of such cards requires inexpensive equipment and errors are easily 
corrected. However, sorting such cards requires much physical manipula¬ 
tion and the handling of 5,000 or more cards with needles becomes a real 
chore. The number of notch or code positions is limited by the size of the 
card and the larger the card the more difficult it is to handle. Finally, such 
cards are expensive to prepare and reproduce and are subject to rather rapid 
deterioration with much use. 

Machine-Sorted Cards. The chief advantages of this type of card are 
the greatly increased coding possibilities on small cards and the ease and 
accuracy with which they may be handled. On a 3% by inch IBM card 
there is space for 960 direct code punches. It would take a notched card over 
four feet square to provide the same information. Also, machine sorting is 
so effortless that tens of thousands of cards pose no special problems. IBM 
cards are inexpensive and easy to prepare and reproduce. On the other 
hand, only a limited amount of information can be printed on such cards. 
It thus becomes necessary to use the cards merely to index the standard 
data rather than to provide a means of recording and filing such data. This 
necessitates the maintenance of two files. The file of detailed standard data 
may be kept in whatever form it is obtained, i.e., film strips, recording 
charts, published curves or tabulations; the actual searching for wanted 
portions of this data is done with punched cards which bear serial number 
references to the location of the standard data. IBM cards necessitate 
the use of mechanical equipment. While this can be quite expensive in some 
cases, most of the applications described in this chapter require only a 
sorter. 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 177 


It has been widely agreed that machine sorting is best for universal 
searches where many thousands of cards are involved and the data are 
complex, and that hand sorting is most efficient for small numbers of cards 
indexing limited ranges of data. Many laboratories find it convenient to use 
both types. In this discussion that follows, examples will be presented 
showing how both hand-sorted and machine-sorted cards are applied in 
handling data in several fields of physical analytical chemistry. 

Infrared Absorption Spectroscopy 

The physical data obtained from a compound by an infrared absorption 
spectrograph are complex and are usually represented by a plot of the per 
cent transmittance versus the wavelength or frequency of the infrared 
radiation. Modern spectrographs produce such a plot or spectrogram auto¬ 
matically but the shape and size of the spectrogram varies with the make 
of the instrument. Much data is still hand plotted for publication. Here, 
then, is a case where standard data obtained from compounds of known 
purity, with which the unknown spectrogram must be compared for identi¬ 
fication purposes, exist in the form of thousands of spectrograms, plotted 
in several coordinate systems, in a variety of shapes and sizes, and located 
in many books, 1 ' 3 • journals and catalogs. 4 6 • * A This makes the physical 
matching of standard and unknown data for qualitative analysis unduly 
time-consuming. Consequently, a great deal of effort has gone into the 
problem of adapting punched card techniques to the solution of the cor¬ 
relating problem. 

The fact that much of the data in an infrared spectrogram can also be 
correlated with specific structural features of the compound from which it 
was obtained has made it desirable to include in punched card systems the 
details of chemical structure or groups which are singly responsible for spe¬ 
cific segments of the spectrogram. In order to anticpate future correlations 
of structure and absorption data, the chemical structure codes usually 
have been as embracing as space limitations on the various cards permit. 
In some systems it has been desirable to include important elements, melt- 

1 Barnes, R. B., R. C. Gore, U. Liddle, and V. Z. Williams, “Infrared Spectro¬ 
scopy,” New York, Reinhold Publishing Corp., 1944. 

1 Dobriner, K., E. R. Katzenellenbogen, and R. N. Jones, “Infrared Spectra of 
Steroids,” New York, Interscience Publishers, 1953. 

* Randall, H. M., R. G. Fowler, N. Fuson, and J. R. Dangl, “Infrared Determina¬ 
tion of Organic Structures,” New York, D. Van Nostrand Company, 1949. 

4 “Catalogue of Infrared Spectral Data,” American Petroleum Institute, Re¬ 
search Project 44, Carnegie Institute of Technology, Pittsburgh, Pennsylvania. 

• “Catalogue of Infrared Spectrograms,” Samuel P. Sadtler and Son, Inc., 1517 
Vine Street, Philadelphia 2, Pennsylvania. 

•National Research Council Committee on Spectral Absorption Data, Mr. J. 
J. Coraeford, Secretary, National Bureau of Standards, Washington 25, D. C. 

•a “The DMS System,” London, Butterworths Scientific Publications, 1956. 



178 


PUNCHED CARDS 


ing or boiling points, and other physical characteristics of the compounds 
in the information coded into the cards. This provides greater flexibility in 
card sorting when such data are available. 

Hand-Sorted Cards. A great many laboratories have developed codes 
for and make use of such cards to meet their particular needs. It will not 
be possible to describe all of these. However, many of them are quite similar 
since the objective in all cases is much the same. The system currently 
distributed by the National Research Council Committee on Spectral 
Absorption Data in collaboration with the National Bureau of Standards 
represents the best features of some eighteen such systems. It was developed 
by a Punched Card Committee appointed by the 1948 Symposium on 
Molecular Structure and Spectroscopy at Ohio State University. Because 
of the wide acceptance of this card and system, the description which 
follows is sufficiently complete to provide working instructions for the use 
of the cards. 

The edges of the N.R.C.-N.B.S. card (Figure 9-1) are divided into four 
fields which provide for coding and sorting the following data: 

(1) Positions of Major Absorption Bands 

(2) Melting or Boiling Point 

(3) Molecular Functional Groups 

(4) Number of Carbon Atoms 

Fields (1) and (2) permit entry to the file in the case of unknown compounds 
while (3) and (4) permit entry in the case of known compounds. 

(1) Wavelength-Wave Number. This field includes from 2.70 (3700 recipro¬ 
cal centimeters) to 40.0 microns (250 reciprocal centimeters) in nonlinear 
intervals. The spectral range covered by each notch position is printed on 
the card and the coding of each major band is achieved by notching the card 
between the printed values which encompass the major band. Only the 
relatively strong absorption bands of a given spectrogram are coded into 
the card, and sorting these corresponding fields serves to isolate those cards 
indexing compounds whose spectra match the one sought for. A compari¬ 
son of the unknown spectrogram with those appearing on the back of the 
card (Figure 9-2) serves to establish the identification. 

(2) Melting or Boiling Point. A melting or boiling point is notched into 
the field on the right of the card. The familiar 1,2,4,7, SF system for notch¬ 
ing numbers is used and space is provided for three-digit, whole-number 
values. A small field adjacent serves to indicate whether the notched value 
is a melting or a boiling point as well as to indicate negative values when 
necessary. 

(3) Molecular Functional Groups. Holes numbered 1 through 31 are a 
direct sorting code for molecular functional groups, which were chosen be¬ 
cause they are known or suspected to have characteristic absorption in the 




Figure 9-1. Front of NRC-NBS Compound Card. 











































































180 


PUNCHED CARDS 


infrared region. Both deep and shallow notching is used in this field and 
abbreviations of the corresponding molecular group are printed under each 
hole on the card. Table 9-1 gives a complete description of each abbrevi¬ 
ated term. 

(4) Number of Carbon Atoms. This field provides for recording such in¬ 
formation about the compound with the 1, 2, 4, 7, SF code and sufficient 
space is available to indicate 39 or less atoms. This information, together 
with sorts on functional groups, serves to help isolate the card indexing a 
given, known compound and to eliminate unwanted cards when partial 
information is available about an “unknown” compound. 

Space is provided on the face of the card for the name of the compound, 
structural and empirical formulas, range of the spectrogram (printed on the 
back of the card), state of the sample, purity, serial number, etc. A tabular 
list of references to other published spectra of and information about the 
particular compound is also provided. The back of the card carries the 
infrared absorption spectrogram of the compound and a list of the spectral 
positions of the absorption bands. These cards are available from the Na¬ 
tional Bureau of Standards. 8 

Mention should be made of a few other notched card systems which have 
worked quite well. One such system has been developed by the Shell 
Development Company. 7 In order to provide more notch positions, a large 
card, 8^2 by 11 inches, is used. Into this card, by means of direct codes, 
the spectral region covered, the number and kind of atoms, the number of 
carbon atoms and the number and kind of functional groups are notched. 
Two-number codes are used to notch “functional groups,” atoms, boiling 
point and wavelength positions of major absorptions of the spectrogram 
into the rest of the card. Unique is the use of very small “functional groups,” 
the indication of whether these groups are interacting or independent and 
whether they are acyclic, cyclic or exocyclic with respect to possible rings 
in the compound being coded. Although the use of the two-number code and 
the small “functional groups” adds somewhat to the task of coding and 
sorting the data, it does permit greater resolution in notching absorption 
positions and structural detail. Pioneers in the development and testing of 
punched card methods for indexing and sorting infrared absorption data 
must include workers at Dow Chemical Company, 8 Rohm and Haas Com¬ 
pany* and United States Rubber Company. 10 The success of their effort 
is reflected in the N.R.C.-N.B.S. card of today. Many others were quick to 
support these efforts and to supply modifications and improvements. Par- 

7 Brat tain, R. R., personal communication, February, 1952. 

8 Wright, N., “Infrared Spectroscopy,” Am. Chem. Soc. Abst., 6L, 108th Meeting, 
New York Cith, (September 1944). 

• Stroupe, James D., personal communication, December, 1952. 

10 Hampton, R. R., personal communication, December, 1952. 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 181 


Table 9-1. List of Abbreviations for National 
Research Council Infrared Punch Cards 




Compound card—functional group code 

No. 

Position 

Code 

Class or Sub-class 

1 

Shallow 

Cl 

All cpds. containing Chlorine. 

1 

Deep 

Cl. 

Cpds. with more than one Cl atom per mole¬ 
cule. 

2 

Shallow 

HAL 

All cpds. containing F, or I. 

2 

Deep 

Br 

All cpds. containing Br. 

3 

Shallow 

CONJ 

All cpds. with conjugated nonaromatic double 
bonds. 

4 

Shallow 

C=C 

All cpds. with an acetylenic bond. 

5 

Shallow 

HC 

All hydrocarbons, incl. C isotopes, but not H 
isotopes. 

5 

Deep 

SAT 

All saturated hydrocarbons, incl. C isotopes, 
but not H isotopes. 

6 

Shallow 

POL 

All polymers. 

6 

Deep 

CO-POL 

All copoly mere. 

7 

Shallow 

C=C 

Cpds. wdth aliphatic C=C double bonds, not 
incl. in 7 deep and 8. 

7 

Deep 

R,C=CHR 

Cpds. containing the aliphatic double bonds of 
types RJtbC=CHR c and RjC == CRj , R 
same or different but not H. 

8 

Shallow 

=CH, 

Cpds. with aliphatic double bonds of the tvpe 
C=CHi . 

8 

Deep 

11,0= 

Cpds. with aliphatic double bonds of the type 
R»RbC=CHj . 

9 

Shallow 

AR 

All cpds. containing an aromatic carbon ring 
not incl. in 9 deep, 10, 11, 12. Fusing with 
another ring constitutes substitution. 

9 

Deep 

MONO 

All cpds. containing a mono-substituted aro¬ 
matic carbon ring. 

10 

Shallow 

PARA 

All cpds. containing para di-substituted aro¬ 
matic carbon rings. 

10 

Deep 

ORTHO 

All cpds. containing ortho di-substituted aro¬ 
matic carbon rings. 

11 

Shallow 

META 

All cpds. containing meta di-substituted aro¬ 
matic carbon rings. 

11 

Deep 

UNSYM 

All cpds. containing unsymmetrical tri-sub- 
stituted aromatic carbon rings. 

12 

Shallow 

SYM 

All cps. containing symmetrically tri-substi- 
tuted aromatic carbon rings. 

12 

Deep 

VIC 

All cpds. containing vicinally tri-substituted 
aromatic carbon rings. 

13 

Shallow 

HET 

All cpds. containing heterocyclic rings. 

13 

Deep 

HET-N 

Heterocyclic cpds. with nitrogen in the ring. 

14 

Shallow 

ALCY 

Cpds. with carbon rings other than aromatic 
rings. 

14 

Deep 

ALCY-A 

Cpds. with C=C in a carbon ring other than an 
aromatic ring. 



182 


PUNCHED CARDS 


Table 9-1. Continued 

Compound card—functional group code 


No. 

Position 

Code 

Class or Sub-class 

15 

Shallow 

OH 

All cpds. containing OH except C=C(OH); in¬ 




cluding hydrates. 

15 

Deep 

ALC 

All alcohols (excluding phenols). 

16 

Shallow 

AC 

All carboxylic acid anhydrides, and salts. 

16 

Deep 

ACII) 

All carboxylic acids. 

17 

Shallow 

AL 

All aldehydes. 

17 

Deep 

KE 

All ketones. 

18 

Shallow 

EST 

All esters of carboxylic acids. 

18 

Deep 

ACET 

Esters of acetic acid. 

19 

Shallow 

ETH 

All oxygen C—0—C ethers. 

19 

Deep 

METH 

Methyl oxygen ethers. 

20 

Shallow 

CO 

Cpds. with C-to-0 bonds not included in No. 15 




Nos. 19 and 20 deep. 

20 

Deep 

AMD 

Acid amides and N-substituted acid amides. 
—C=0(NH 2 ), —C=0(NHR), —C=0(NR 2 ). 

21 

Shallow 

CS 

All cpds. with C—S bonds. 

21 

Deep 

c=s 

All cpds. with C=S bonds. 

22 

Shallow 

s 

All other cpds. containing sulphur not in 21 




and 22 deep. 

22 

Deep 

so 

All cpds. containing S-to-0 bonds including 




inorganic radicals. 

23 

Shallow 

CN 

All cpds. with N-to-C bonds, including ter¬ 




tiary amines but not 20 deep, 23 deep and 24 
deep. 

23 

Deep 

C=N 

Nitriles. 

24 

Shallow 

NH 

Secondary amines, not acid amides. 

24 

Deep 

AM IX 

Primary amines. 

25 

Shallow 

NO 

All cpds. with N-to-O bonds, except 25 deep. 

O 

y 

25 

Deep 

NO, 

All cpds. having N groups. 




\ 

0 

26 

Shallow 

X 

All cpds. containing nitrogen not covered else¬ 




where. 

26 

Deep 

X—X 

Cpds. containing N-to-N bonds. 

27 

Shallow 

SI 

All cpds. containing silicon. 

27 

Deep 

— 

Not assigned. 

28 

Shallow 

XOX-M 

All cpds. which contain non-metals not covered 




elsewhere. 

28 

Deep 

P 

Cpds. containing phosphorus. 

29 

Shallow 

MISC 

Miscellaneous cpds. not included elsewhere. 

29 

Deep 

MET 

Organo-metallic cpds. 

30 

Shallow 

1SOT 

Cpds. containing isotopes other than D and T 

30 

Deep 

DECT 

Cpds. containing deuterium and tritium. 

31 

Shallow 

— 

Unassigned. 

31 

Deep 

— 

Unassigned. 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 183 


ticular mention should be made of a unique approach to the problem made 
by a worker at Cornell Medical College, 11 where the system devised for 
cataloging x-ray diffraction data based on the three strongest lines 12 was 
applied to infrared absorption data. A new and comprehensive system for 
coding chemical structure and spectral absorption data into edge-notched 
cards has been proposed by Thompson 12A which forms the basis of an ex¬ 
tensive catalogue of data being issued from England. 6A These data are 
being coded into IBM cards also. 

Machine-Sorted (IBM) Cards. An early application of IBM cards and 
equipment to the problems of sorting and correlating infrared absorption 
data for purposes of qualitative analysis was developed by Wyandotte 
Chemicals Corporation. 13 The American Society for Testing Materials* has 
since assumed the responsibilities connected with the development and 
maintenance of this and other indexing systems originating at Wyandotte, 
and which are being used by an increasing number of laboratories. This 
provides for the perpetuation and orderly modification of the systems in the 
interests of a greater number of people, provides a mechanism for supplying 
the cards and codes through a single agency, and furnishes the manpower 
to insure greater coverage and accuracy of the data issued. A research fel¬ 
lowship at the National Bureau of Standards has been established by the 
American Society for Testing Materials to provide for the preparation of 
master cards coded by members of the Standard Data Subcommittee of 
A.S.T.M. Committee E-13. A description of the Wyandotte-A.S.T.M. 
system for indexing infrared absorption and chemical structure data, to¬ 
gether with pertinent codes and charts, follows; other system will be dis¬ 
cussed later in this chapter. 

Wyandotte-A.S.T.M. infrared punched cards are designed to facilitate 
the sorting of spectral absorption data for the purpose of matching spectro¬ 
grams in qualitative analysis and to aid in correlating chemical structure 
and absorption band positions. The cards are merely an indexing system 
which enables one to make rapid, accurate searches by machine and inex¬ 
pensively to obtain and maintain ready reference to all published spectra. 

11 Clark, C., “Cataloguing of Infrared Spectra,” Science, 111, 632-633, (June 9, 
1950). 

11 Hanawalt, J. D., H. W. Rinn, and L. K. Frevel, “Chemical Analysis by X-Ray 
Diffraction,” Ind. Eng. Chem., Anal Ed., 10,457 (1938). 

** A Thompson, H. W., “The Documentation of Molecular Spectra,” Journal of the 
Chemical Society, pages 4501-4509, (1955). 

** Kuentzel, L. E., “New Codes for Hollerith Type Punched Cards,” Analytical 
Chemistry, 23, 1413-18 (1951). 

* Thanks are due the American Society for Testing Materials for permission to re¬ 
produce in this chapter its copyrighted Codes and charts pertaining to ASTM- 
Wyandotte punched cards for indexing spectral absorption data. 



184 


PUNCHED CARDS 


Through the cooperation of the Technical Information Division of the 
Battelle Memorial Institute, the National Bureau of Standards and mem¬ 
bers of A.S.T.M. Committee E-13, all publications are monitored for infra¬ 
red spectra which, together with those issued by regular publishers of 
spectra, are coded and incorporated into the card files. Cards, indexes and 
instruction booklets 14 may be obtained from the A.S.T.M. 

For reason of simplicity and economy of sorting time, the system makes 
widest possible use of direct codes and requires the use only of a standard 
Type 82, 80-1 or 80-2 Sorter available from International Business Ma¬ 
chines, Incorporated. Although considerable information about the opera¬ 
tion and use of common pieces of IBM equipment is available elsewhere 
in this book, the manipulation of the sorter as it applies to the handling of 
these cards will be reviewed briefly. 

The IBM card bears eight vertical columns of numbers from 0 through 9 
which mark the positions where small rectangular holes will be punched to 
record data. Above the 0 position in each column are two unmarked over¬ 
punch positions referred to as “x” and “y” positions, with the latter being 
uppermost. Numbers can be punched in directly. Thus, the number 457 
can be recorded by punching the digits 4, 5 and 7 in any three adjacent 
columns. Letters of the alphabet require a special two-punch-per-column 
code. When used for these purposes, only one digit or letter can be punched 
into a given column. However, a direct code can be set up to relate any 
given punch position to any particular item. Thus, a punch at number 7 in 
column 42 can always mean that there is an —OH functional group in the 
compound described by the card. With this type of coding it is possible to 
record from one to 12 items independently and at the same time in a single 
column of a card. Moreover, by making use of the selector switches, any 
item so coded may be searched for singly regardless of the number of other 
items coded into the same column. In sorting operations, the sorter senses 
one column at a time. As the cards pass over a metal cylinder, a small 
metal brush makes electrical contact with the cylinder through the holes 
punched in the cards. This causes the card to be shunted into a pocket cor¬ 
responding to the number at which the punch is made. If there is more than 
one punch in the column, the machine sends the card to the pocket bear¬ 
ing the highest number so punched. Positions “x” and “y” rate below 0, 
with the “y” position being the last sensed by the machine. However, if 
any one or more of the 12 selector switches are moved to the “off” position, 
the machine will ignore punches at those positions and respond to any re¬ 
maining punches, and if there are none, it will then send the card to the 

14 “Codes and Instructions for Wyandotte-A.S.T.M. Punched Cards Indexing 
Spectral Absorption Data,” the American Society for Testing Materials, 1916 Race 
Street, Philadelphia 3, Pennsylvania, 1954. 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 185 



Figure 9-3. Wyandotte-A.S.T.M. Infrared Data Card. 


“reject” pocket. Sorting for cards that do not have punches at particular 
positions can be as useful as sorting for cards that do have punches; the 
former has been termed “negative” sorting while the latter is referred to as 
“positive” sorting. These operations are simple to carry out and examples 
described in the chapter will provide additional details. 

The Wyandotte-A.S.T.M. card indexing infrared absorption data (see 
Figure 9-3) is divided into the following areas for coding purposes: 

(1) Infrared Absorptions—columns 1 through 28 

(2) Chemical Classification—columns 32 through 57 

(3) Semi-empirical Formula—columns 58 through 62 

(4) Melting or Boiling Point—columns 63 through 65 

(5) Reserved by A.S.T.M.—columns 29 through 31 

(6) Reserved for Private Use—columns 66 through 70 

(7) Reference or Serial Number—columns 71 through 80 

All the data available from one compound and covered by the codes are 
punched into one card. In order that this description of the card and sys¬ 
tem may serve to instruct one in the use of the cards, codes and examples 
are included. 

(1) Infrared Absorptions. The coding of absorption band positions is done 
in terms of wavelength in microns. As a general rule, all bands having an 
absorbance ratio with the strongest band in the spectrogram of 1:10 or 
more are coded. From columns 1 through 25, the column number is taken 
as the whole number value of the absorption band and the fractional part 
is rounded off to the nearest tenth of a micron and coded by punching the 
digit into the appropriate column. Thus, a band at 7.38 microns would be 
coded by a single punch at the 4 position in column 7. From 25 to 50 mi¬ 
crons the punching resolution is 1.0 micron and coding is achieved by punch¬ 
ing the units values into columns headed by appropriate tens values. Thus 
a value of 34.8 microns is coded by a single punch at the 5 position in column 


































186 


PUNCHED CARDS 


27 which is headed by the number 30. Punches at the “x” position in this 
section of the card serve to indicate that there are no data available for the 
region covered by the particular column or columns involved. Finally, a 
“y” overpunch is used to code the position of each very strong band which 
may be expected to persist in the spectrum of a considerably diluted sample 
of the material. 

Sorting operations for coded absorption data follow one of two possible 
courses depending whether the unknown material is relatively pure or is a 
mixture of several compounds in roughly equal concentrations. In the 
former case, positive sorting on the strong band positions is appropriate 
while the latter case involves initial negative sorting before making positive 
sorts. Each method is described briefly. 

Positive sorting on coded absorption band positions seeks to segregate 
cards bearing code punches corresponding to those of the major absorptions 
in the spectrogram of the unknown material. The band code positions are 
sorted for one at a time, beginning with the most characteristic or unique 
band. This tends to eliminate the greatest number of cards in the initial 
sort and is usually achieved by sorting on the longest wavelength band first. 
By making use of the selector switches, the sorter can be set to search for 
cards having either specific individual punches or a range of consecutive 
punch positions. The latter operation permits sorting for a narrow range 
of absorption band positions simultaneously when the exact location of the 
band peak is not apparent in the spectrogram and thus avoids missing the 
wanted card. Each sort is made upon the small residue of the previous sort 
and the number of cards diminishes rapidly so that a single card frequently 
results in from three to five sorts even when starting with several thousand. 
Each card bears a serial number which directs the searcher to the location 
of the standard infrared spectrogram from which identification of the 
unknown spectrogram may be made. 

In spectra of mixtures of compounds, it is not known which bands belong 
to chemical individuals so that sorting of the type just described is not 
feasible. One seeks first to eliminate all cards which have absorption bands 
in wavelength regions where the spectrogram of the unknown does not 
have any, since none of these materials could possibly be a component of 
the mixture. This is accomplished by negative sorting of the “transparent” 
regions of the unknown spectrogram, i.e., setting the sorter to segregate 
into pockets all cards that do have bands in these regions, then discarding 
them and keeping the cards which accumulate in the reject pocket. Use of 
the “y” overpunches facilitates negative sorting since the relatively weak 
bands of a minor component do not interfere with such sorts in otherwise 
transparent regions. The relatively small deck of cards that results from 
the negative sorting operation is then subjected to positive sorting on the 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 187 


band positions, following a systematic trial and error schedule that con¬ 
siders the possible combinations of bands that may characterize an in¬ 
dividual component of the mixture. Here, again, it is expedient to begin 
with the long wavelength bands and test-sort all possible combinations un¬ 
til an identification is made. When this is done the bands belonging to that 
compound are eliminated from consideration and the process continued on 
the remaining bands. Although it sounds complicated, the actual time in¬ 
volved is small and the process is easy to carry out. A typical example of 
the analysis of a three-component mixture starting with over 3,000 cards 
required about 18 minutes of negative sorting and 3 minutes of positive 
sorting on the residue deck to identify the three unknowns. 

(2) Chemical Classifications. The philosophy behind the development and 
use of the Chemical Classification Code attempts to divorce the complexi¬ 
ties of the names and chemistry of organic compounds from the codes and 
coding operations used to characterize them. It is not intended that each 
such characterization be unique for each different molecule since the pur¬ 
pose of the code is to provide a means of segregating compounds into re¬ 
lated groups. Coding is based entirely upon a detailed structural formula 
and a recognition of the “code units” which make up the formula. These 
code units, in many cases, are the same as familiar reactive groups or radi¬ 
cals that enter into the chemistry and naming of organic compounds, but 
such names and chemistry as may be associated with the code units must 
not restrict the use of the unit wherever applicable under the rules pre¬ 
sented. 

Table 9-2 relates the code punch positions on the card in terms of column 
and row numbers to the coded items of structural features used to charac¬ 
terize compounds. References to these punch positions are made by giving 
the column number, then a dash followed by the punch positions. For ex¬ 
ample, 32-0, 2, 4 indicates punch positions 0, 2 and 4 in column 32 which 
code the presence of elements oxygen, sulfur and chlorine, respectively. 
Table 9-3 lists some examples of the types of compounds which the code 
units index. Following is a column-by-column discussion of the chart and a 
presentation of the rules set up to insure uniformity in coding. 

Column 32 provides for the coding of the identity of elements commonly 
found in organic compounds. Carbon and hydrogen are not coded directly 
but hydrocarbons are indicated when there are no punches in this column. 
In applying the code a punch is made in the proper position for each dif¬ 
ferent element regardless of the number of such elements in the compound. 
The coding of less common elements is provided for in columns 56 and 57. 
A punch at the “y” position in column 32 must accompany any coding in 
these latter columns. 

Column 33 codes the type and location of unsaturated carbon-to-carbon 



188 


PUNCHED CARDS 


Table 9-2A. 


| Row 

Column 32 
Elements 

Column 34 
Structure 

Row 

WmnSSmniREm 

Column 38 
Miscellaneous 

Row 

D 

0 

Acyclic 

0 

12 or more 

Solid 

0 

n 

N 

Alicyclic 

a 

1 

Liquid 

i 


S 

Aromatic 

2 

2 

Gas 

2 

3 

F 

Heterocyclic 

3 

3 

Organo-metallic 

3 

D 

Cl 

Fused Alicyclic 

D 

4 

Isotopic 

4 

5 

Br,l 

Fused Aromatic 

5 

5 

Indeterminate 

5 

6 

P, Bi 

Fused 

Heterocyclic 

6 

6 

Solution 

6 

D 

As,Sb 


D 

7 

Polymer 

7 

8 

Si.Ge 

ini ttm 

8 

8 

Chelate 

8 

9 

Sn.Pb 

5 Member Ring 

9 

9 

Hydrate 

9 

D 

B,A1 

6 Member Ring 

D 

10 

KBr Plate 

X 

D 

Other 


a 

1 1 


Y 


Row 

Column 33 
Unsaturation 

Column 35 
Rings— Chains 

Row 

Column 37 
Substitutions 

Column 39 
Miscellaneous 

Row 

0 

Ring 

Rings 

D 

[mono] 

cis 

0 

1 

1 

1 

a 

1 D.2] 

trans 

1 

2 

2 

2 

2 

2 

[1.3] 

spiro 

2 

3 

3 

3 

3 

3 

0.4] 

dextrorotory 

3 

n 

4 

4 

D 

4 

[', 2 , 3 ] 

levorotary 

4 

5 

5 

5 

a 

5 

0,2.4] 

symmetrical 

5 

6 

6 

6 

6 

6 

0.3.5] 

unsymmotricol 

6 

7 

7 

7 

a 

7 

0 , 2 , 3 , 4 ] 

vicinal 

7 

8 

8 

8 


8 

[', 2 , 4 , 5 ] 

Salt 

8 

9 

9 

9 

a 

9 

0.2 , 3 , 5 ] 

Inorganic Ester 

9 

a 

-c=d- 

10 


10 

[pento] 


X 

tz 

-c=c- 

II or more 

D 

[hexa] 

Inorganic 

Y 


bonds. In every case, except for aromatic unsaturation, the unsaturation 
is coded as to type, that is, double bond or triple bond or both, by punches 
at 33-x or 33-y, or both. Numbers in this column are used to indicate the 
location of the unsaturated bonds subject to the following rules: 

(1) If the unsaturation is located in a ring, then a 33-0 is required. When 























































































































QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 189 



this punch is lacking, it is understood that unsaturation in a chain is being 
coded. 

(2) Unsaturation at positions requiring numbers higher than 9, Greek 
letters, or primed numbers are not coded. 

(3) The use of the position codes is restricted to compounds containing a 


























































































































190 


PUNCHED CARDS 


Table 9-2C. 


Row 

Column 50 
N-S 

vSlfll 

[ Row 

yV -' j: 

kctsIhB 

1 

1 IBnHHdHHflrti- 

Row 

0 


-0C(=S)S- 

-SC(=0)S- 

D 


Se,Te, Po 

0 

1 

^NC(=S)S- 
— SC(=ltl)S— 

-OCbSX)- 

-0C(=0)S- 

n 


Ga.ln.TI 

a 

2 

-C(=S)< 

-C(=!<I)S- 

-C(=S)0- 

-C(=0)s- 

2 

-oc(«A)s- 

Zn,Cd,Hg 


3 


-S(0*)0- 

3 

-S(=0)NC 

Cu, Ag,Au 

[Jl 

4 

— SCN 

-S(=0)0- 

a 

iNS(0t)NC 

Fe.Co.Ni 

D 

5 

— NCS 

-S(=0)S- 

-S(--S)0- 

5 

^NS(0s)0- 

Cr,Mo,W,U 


6 

JTNSNC 


6 

^NS(Ot)- 

V,Cb,To,Po 

"H 

7 

= NS- 
^NS- 

^S0* 

a 

^NS(=0)0- 

Ti.Zr, Hf.Th 

a 

8 

-N=S 

^s=o 

8 

-NS0 

Sc.Y.Lo.Ac 

8 

9 

^S:N- 

-oso- 

9 


Ru,Rh,Pd, 

Os. Ir. Pt 

9 

X 

Other 

other 

a 

Other 

R.E. 

D 

Y 

Heterocyclic 

Heterocyclic 

D 

Heterocyclic 

Heterocyclic 

D 


Row 

Column 51 

N-S 

Column 53 

o-s 

Row 

Column 55 
N-O-S 

Column 57 
Elements 

Row J 

0 


-0S(0t)0- 

D 


Li 

D 

i 


-0S(=0)0- 

n 


Na 

a 

2 



2 


K 

2 

3 



3 


Rb.Cs 

3 

4 



D 


Be 

D 

5 



5 


Mg 

5 

6 



6 


Ca 

6 

7 



a 


Sr, Bo 

a 

8 



8 



8 

9 



9 



9 

X 

Conjugated 

Conjugated 

a 

Conjugated 

Conjugated 

a 

Y 



a 


Other 

a 


single chain, a single ring or a fused ring system where the Geneva System 
for chains or the Patterson Ring Index can be applied without ambiguity. 

(4) Unsaturation in benzene rings, fused or otherwise, or in alicyclic 
rings as a result of fusion with aromatic rings is not coded here. 

(5) Where both cyclic and chain systems are present in a single compound 















































































Table 9-3. Examples of the Types of Compounds Coded by the Code Units in 
the Chemical Classification Code Chart 
Following are examples of the types of compounds which the various code units in 
the chart may index. It is to be understood that these examples do not restrict the use 
of the code units in the indexing of other types of compounds in which they may ap¬ 


pear. 

42-0 acids 

42-1 esters, salts, lactones anhydrides 

42-2 aldehydes 

42-3 ketones 

42-4 carbonates 

42-5 ortho carbonates 

42-6 ortho carboxylates 

42-7 alcohols, phenols 

42-8 ethers, oxy compounds 

42- 9 peroxides 

43- 0 oxonium compounds 

43-1 ozonides 

43- 2 acetals 

44- 0 amidines 

44-1 guanidines 

44-2 nitrilo or cyano compounds 

44-3 isonitrilo compounds 

44-4 primary amines 

44-5 secondary amines 
44-6 tertiary aminds 
44-7 imines 

44-8 hydrazones, hydrazines 

44- 9 azo or diazo compounds 

45- 0 triazenes 

45-1 diazonium compounds 

45-2 quaternary ammonium com¬ 
pounds 

45-3 ammonium compounds 

45-4 cyanamides 

45- 5 triazo compounds, azides 

46- 0 thionothiolic compounds, carbodi- 

t hi oates 

46-1 thioaldehydes 

46-2 thiones, thioketones 

46-3 trithio carbonates 

46-4 thiols 

46-5 sulfides 

46-6 disulfides, polysulfides 
46-7 sulfonium compounds 
46-8 perthio compounds 


48-0 carbamyl compounds, carbamates 

48-1 ureido compounds 

48-2 amides, imidic compounds, lactams 

48-3 isocyanates 

48-4 cyanates 

48-5 nitro amines 

48-6 nitroso amines 

48-7 azoxy compounds 

48-8 nitrates 

48- 9 nitrites 

49- 0 nitro compounds 

49-1 nitroso compounds 

49-2 isonitroso compounds, oximes 

49- 3 amine oxides 

50- 0 thiourido compounds 

50-1 thiocarbamyl compounds 

50-2 thioamides, thiomides 

50-3 

50-4 thiocyano compounds 

50-5 isothiocyano compounds 

50-6 diamino sulfides 

50-7 sulfimes, sulfenamides 

50-8 sulfamino and sulfinyl compounds 

50-9 sulfilimines 

52-0 dithiocarbonates 

52-1 thiocarbonates 

52-2 thiolic, thionic compounds, carbo- 
thioates 
52-3 sulfonates 
52-4 sulfinates 
52-5 thiosulfinates 
52-6 thionates 
52-7 sulfones 

52-8 sulfoxy compounds, sulfinyls 

52- 9 sulfenates 

53- 0 sulfates 

53- 1 sulfites 

54- 0 thiocarbamates 

54-1 carboxamido sulfides 
54-2 

54-3 sulfinamides 
54-4 sulfamides 
54-5 sulfamates 

54-6 sulfonyl amines, sulfonamides 
54-7 amino sulfinates 
54-8 sulfinyl amines 
54-9 
191 



192 


PUNCHED CARDS 


and unsaturation is present in only one or the other, it is coded as to loca¬ 
tion. 

(6) Where both cyclic and chain systems are present in a single compound 
and both contain unsaturation, the position code is applied to the largest 
ring or fused ring system. 

Column 34 codes the major structural features of a compound and is 
largely concerned with the type and size of rings. The use of these codes in 
describing a molecular structure is governed by the following rules: 

(1) An “acyclic” code is used whenever an open chain of two or more 
atoms other than hydrogen form a part or all of the molecule or whenever 
the molecule consists of only one atom other than hydrogen. Carbon atoms 
in rings are not counted as part of chains. Thus, methane, ammonia, phenyl- 
hydrazine, and ethylbenzene would require 34-0 punches, but toluene, 
phenol and aniline would not. 

(2) Each individual type of ring present in a single molecule is coded by 
a single appropriate punch. Each member of a fused ring system is coded 
separately if different types are involved. All rings other than aromatic 
(benzene) and heterocyclic are considered alicyclic and only benzene rings 
are coded aromatic. 

(3) No portion of any ring, except that involved in fusion, is coded more 
than once. Thus, multiple ring systems formed by bridging are individually 
coded but the enveloping ring is not. 

(4) The size of aromatic rings is not coded with a 34-x punch. 

Column 35 provides for coding the length of carbon chains or the number 

of rings in a compound. Use of the following rules will insure uniformity of 
coding: 

(1) If there is only one ring, or if there are no rings in the compound, 
the length of the longest normal earbon-to-carbon chain is coded by an ap¬ 
propriate punch. One carbon atom is considered a “chain,” but carbon 
atoms in rings are not to be counted as part of such chains. 

(2) If there are two or more rings in the compound and aromatic rings 
are involved, both the total number of rings and the number of benzenoid 
rings are punched into column 35 together with a punch at 35-0. In any 
case, the total number of rings is punched in. Each ring in a fused ring, 
spiro or bridged system is counted separately. Rule 3 under column 34 
also applies here. 

Column 36 codes the total number and the number of different kinds of 
“code units” observed in the structure of the molecule. Both numbers 
should be coded into this column when there is a difference between the 
total number and the number of different kinds. The following rules assist 
in arriving at proper totals: 

(1) Consider all code designations covered on the chart by columns 42 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 193 


through 55 except codes for “heterocyclic” and “conjugated.” Each desig¬ 
nation made is a code unit. 

(2) Consider all atoms other than C, H, N, 0 and S. Each other element 
counts as a code unit. 

(3) For the total number of code units, count each unit as many times 
as it appears in the structure. 

(4) For the number of different kinds of code units, count each type once. 

Column 37 provides for locating the positions of substituent groups of 

“code units” in a limited number of cases. It is intended that this column 
provide a means of differentiating empirical isomers and is not rigorously 
applied in coding all compounds. One should not attempt to code substi¬ 
tutions in compounds where there is ambiguity as to just what is substi¬ 
tuted on what. The following rules apply: 

(1) Substitution positions requiring numbers higher than 10, or the use 
of Greek letters or primed numbers are not coded here. 

(2) Except as provided in rule 4, below, use of the code is restricted to 
indicating substitution positions on a single carbon chain, or a single ring, 
or fused ring system where application of the Geneva System for chains 
and the patterson Ring Index for cyclic compounds can be made without 
ambiguity. 

(3) In monocyclic compounds which also have acyclic components, code 
the location of substitutions on the ring. 

(4) In polybenzenoid compounds not involving fusion, code designations 
within the brakets (on the chart) are used to indicate the degree and loca¬ 
tion of substitution on the several rings. 

(5) The location of heteroatoms in heterocyclic rings are not to be made 
with this code. 

Columns 38 and 39 code miscellaneous information about the compounds. 
For the most part they are self-explanatory, but the following interpreta¬ 
tions should be made: 

(1) Punches at 38-0, 1, 2 and 6 are used to indicate the physical state 
of the compound both at the time it is analyzed in the spectrometer and 
at room condition. Thus, 38-0, 6 indicates the material to be a solid at 
room temperature but was analyzed in solution. 

(2) Punches at 39-5, 6 and 7 are not to be applied to coding trisubstitu¬ 
tion on benzene or other cyclic compounds but rather to describe the ar¬ 
rangement of heteroatoms in heterocyclic rings such as the triazines where 
substitutions play no part in determining the use of the terms. Such appli¬ 
cation is not limited to rings containing one kind of heteroatom and the 
code may be applied to both five- and six-member rings. 

Columns 40 and 41 code the smaller groups involving carbon and hydro- 



194 


PUNCHED CARDS 


gen only. The following rules apply: 

(1) Code each unit that is observed in the structural formula of the com¬ 
pound being indexed. 

(2) Use the largest code unit that will characterize a group and do not 
code the smaller parts of such a group. Thus, if a —C 2 H 6 group is present, 
code a 40-1 but do not code a —CH 3 at 40-0. 

(3) Under 41-x code all conjugated double bond systems involving car¬ 
bon only, except purely aromatic conjugation. Do code conjugated carbon- 
carbon systems involving a benzene ring if there is at least one double bond 
outside of the ring or if two or more benzene rings form a part of a system. 

(4) Always code the largest system, then do not code any of its parts. 

Columns 42 through 55 provide for coding unit groups involving oxygen, 

nitrogen, or sulfur, singly or together, and with or without a single carbon 
atom. They are arranged in columns depending on the manner in which the 
atoms are involved. The following rules assist in the application of the code: 

(1) Code each unit that is observed in the structural formula whether it 
is part of a ring or not. The only criterion is that the particular arrangement 
of atoms be present. 

(2) Use the largest unit that will characterize a group and do not code 

\ 

any smaller parts. Thus, if the group NC(=0)0— is present, characterize 

/ 


it by a code of 48-0 and do not code 42-3, 42-8 or 44-6. Likewise, if the 
bonds of this code unit were satisfied with hydrogen atoms one would not 
use codes of 44-4 and 42-7 or 42-0. 

(3) Code larger groups than appear in the chart, or those involving two 
or more carbon atoms, by using the least number of largest code units that 
are in the chart. Strict application of this rule is essential regardless of 
one’s feeling for the chemistry or naming of compounds. In some cases this 
rule will necessitate the coding of an atom or two in each of two adjacent 
code units. Thus, for example, in CH 2 =NNHC(=S)NHNH 2 one can ob- 

\ / \ / 


serve the following code units: C=N—, =NN , XC(=S)N , 



XX , C=S, XII, and —XH 2 . The problem is to include all of 

/ \ / / 

these structural arrangements in as few code units as possible. One begins 
by selecting a central carbon atom and observing the greatest number of 
heteroatoms attached to it. In the above example this process yields the 

\ / 


code unit XC(=S)X or 50-0. The other carbon atom calls for a code of 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 195 


\ / 

44-7 and all that remains are the two NN groups which require a 

/ \ 

code of 44-8. Any other possible choice of code units would not involve the 
largest units and any further breakdown would not involve the least num¬ 
ber of units. It will be noted that some of the nitrogen atoms were used 
twice in this process. 

(4) Conjugated double bond systems involving the elements listed at the 
head of each column are coded with the appropriate “x” punch. A conju¬ 
gated double bond system consists of a complete series of alternate double 
and single bonds which may extend through one or more benzene rings. 
One should code only the largest of any such system with the “x” that iden¬ 
tifies the elements involved and not code any of its parts. 

(5) The presence of heterocyclic rings involving elements listed at the 
head of the columns in the chart are coded by “y” punches in the appro¬ 
priate columns. If two or more heteroatoms are involved in a single ring, 
use only the code that involves all of the atoms, except when one of the 
heteroatoms is other than N, 0 or S then a code of 56-y applies. Thus, a 
heterocyclic ring involving both O and S is coded as 52-y and codes of 42-y 
and 46-y are not used. 

(6) Organic salts, including amine salts, etc., are to be coded in the un¬ 
ionized form and a code at 39-8 assigned to indicate a salt is involved. 
Organometallic codes are used only if there is a metal-to-carbon bond in¬ 
volved. 

Columns 56 and 57 provide for coding the less common elements found 
in organic compounds. Provision is made to code conjugate systems and 
heterocyclic groups involving such elements and any element not listed in 
the chart can be coded at 57-y. 

Application of the chemical classification code can best be facilitated by 
a study of examples. For these, the reader is referred to the many thousands 
of compounds that have been coded and punched into cards now commer¬ 
cially available. However, in order to provide an opportunity for one to 
obtain a degree of familiarity with the system with a minimum of effort, 
a number of examples are included in Table 9-4. Use of the coding system 
has spread to many areas far removed from infrared where it is desirable to 
characterize compounds according to structure and then to segregate or 
sort various classes and types of compounds independently from the names 
or chemistry associated with the compounds. 

Sorting the cards to segregate compounds having structural features in 
common or for correlating structure and absorption bands involves merely 
an understanding of how the sorter operates, as described earlier, and know¬ 
ing the code positions for the groups of interest. Both negative sorting and 



196 


PUNCHED CARDS 


CO 


X 


O b- 


oo <m 
co co tt 


<N O 
c5 CO 5 


<N 

O CO >> «-h o 

00 05 ^ oo o 
CO CO Tf< TT 


o ^ x 9 X CO 

S 5 § S S 3 


2 

O 

H 

«! 


co 

05 

J 

o 

J 

u 

8 

H 

M 

o 

fa 

o 

C0 

w 

fa 

a 

*< 

w 


x 

CO 


999 9 o 

<N ^ CO 
CO CO CO 


<N X 

H W* ^ 

o o h w 


CO ^ *o 
CO CO CO 


CO CO 


Jg Jg 

CO CO 


’t 

05 

3 

A 

H 


N x « N 

H W ci H 

9 999 c ?9'? 

0* CO Tf IO CO 00 05 
CO CO CO CO CO co co 


o 

o 

A 

a 

£ B 
J§ o 
3 H? 

pH O 


a 

I 

o 


a 

55 

o 

O 
B 
* 
a 

0—0 

/ \ 


I 


1H 

Q- *« rt 


a 

o 


o 
c5 

o 
*C 

o 

T3 

|«-*v 

2 O—a 

J3 ii a 


O 

H 

A 


0 = 0 


a 

55 

a 

<p 

O 


/ 


a 

0 = 55 


<M 


1 


>> 

A 

I QQ 
a \ 

9 0 = 0 


OJ 


a 

o 



. Potassium salt of gumma parachlorophenoxy crotylmorcaptomethyl 32-0, 1, 2, 4, y 40-0 

penicillin 33-2, x 42-1, 

O 34-0,2,6,8,9 44-y 

II 35-0, 1,3 46-5 


QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 197 



« ® « oo © 

® ss 2 s s 

CO CO “ CO 


00 x 


O <N O 

o ci ci o 
^ ^ ^ 


00 00 
9 9 9 

CO 00 05 

co oo co 


X 

co" 

0**0^ 

>H N co" 

© 9 © <N © 

S? a $8 28 

CO CO CO co co 


© 

V X 

9 co" © ^ 

8 838 


© 

CO © 

9 9 *? t 

co cS co co 


X 

© 

of CO © © 

9 9 9 9 9 9 




198 


PUNCHED CARDS 


p X >) 

^ lO X 
-r ^ ^ 


x >> 



X 

oT 

co ^ >> 

X 1C ^ 


r-H o o o x p x 

cifo^iocooCfTfic 

wwwcocow^^^ 





O 

<N CO 

»-H lO i-H CO 


o O o' CO O CO 

I I • • » I I 


CO CO 


1C CO N X C 
CO CO co CO co 



o 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 199 


X CO 

*5 3 


05 

w o N w 

^ >C h N *■*“> 

O O O Cjl H o N H N 

^4«ioh.ooigo 

««weocowco ; ri5 


x 

50 >> 

H X M) H Tf CC 

9 o“ ^ o'« o 9 n 

ssis&ssii 




w 

X 

O 


i« 

33 

O 


S3 



w 

S3 


«5 


. Chloroacetyl chloride 32-0,4 36-2,3 

34-0 37-1, 2 

CH,C1C(=0)C1 35-2 38-1 

42-3 



200 


PUNCHED CARDS 



x 

X «-T K ** K 

o co h h Sp >> w © p © p p 


iO 

V 

^ W C* t*. © 

9 9 9 9 9 9 

CO S S S c? S 



00 





QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 201 


x ^ 

»o 


os 

° O N 
co *<r 


1-1 T* 

T9°? *r 9 *r 

sssi «ssg 

co v co n 1 ^ 


X 

>* oT »o 

1H CD w IO 00 

9 9 o co 9 ^ oo ^ 
S§eSc3cSe8c§T*!3< 


N N CO 

9 9 9 ^ 9 9999 9919 

SdScScSco w S S S CO 0? CO CO 





202 


PUNCHED CARDS 


positive sorting are necessary, and the following illustrates the funadmen- 
tal operations. Each sorting operation that can be performed narrows the 
character of the structure described by the segregated cards and sorting 
continues on residue decks until the precise structure desired is arrived at. 
Thus, if one wished merely to segregate all cards indexing compounds con¬ 
taining nitrogen, a positive sort at 32-1 would be adequate. Such a sort 
means that the number 1 switch is in the “on” position or pushed toward 
the circumference on the switch ring and all the rest are in the “off” posi¬ 
tion or pushed toward the center of the switch ring. (The large red switch 
always remains in the “on” position.) All cards dropping in the number 1 
pocket will then index compounds containing at least nitrogen, and also 
all other elements. If it is desired to segregate cards indexing compounds 
containing nitrogen only in addition to carbon and hydrogen, a second sort 
on the residue deck is necessary. This is a negative sort at 32-1. In this case, 
the number 1 switch is in the “off” position and all the rest are in the “on” 
position. Now, all cards dropping in the reject pocket will index compounds 
having nitrogen only as the heteroatom. These are the two fundamental 
sorting techniques and the number of switches in the “on” or “off” posi¬ 
tions, together with the column being sorted, will determine the types of 
structure, being either eliminated or segregated, as desired. To continue 
the sorting illustrations, one could take the nitrogen only cards and sort 
positively at 34-2 and 5, then take the cards from pockets 2 and 5 and sort 
negatively at 34-2, 5 and 0 to obtain all the cards indexing aromatic com¬ 
pounds only with nitrogen group substitutions either on the rings or side 
chains. The number of rings or the length of side chains could be regulated 
by sorts in column 35. The number and kinds of nitrogen groups could be 
limited by sorts in column 36 and, finally, the particular nitrogen groups 
could be segregated by sorts in columns 44 and 45. Each sorting operation is 
performed on the wanted residue from the previous sort. The sorting combi¬ 
nations are limitless and thus yield a very great variety of structure combi¬ 
nations. The A.S.T.M. instruction booklet 14 contains additional examples of 
sorting operations. 

A different chemical classification code has been provided for inorganic 
compounds. It is punched into the same areas of the card as the organic 
code but a special code punch at 39-y indicates that an inorganic com¬ 
pound is being coded. Therefore, cards indexing organic and inorganic com¬ 
pounds should be separated by the 39-y sort before sorting on chemical 
structure. In early issues of the infrared cards this inorganic code punch 
was located at 26-0. Otherwise, all infrared cards are identical and may be 
sorted together. The inorganic code is the same one used to index com¬ 
pounds whose x-ray diffraction powder patterns are in the A.S.T.M. files. 
The Elements Code, Table 9-5, provides for all elements with a direct 
punch code for each; the Radicals Code, Table 9-6, supplies additional 





Table 9-5. 

Elements Code 

(43) 

* 32-0 

Actinium—Ac 

36-1 Nickel—Ni 


32-1 

Aluminum-Al 

36-2 Nitrogen—N 


32-2 

Americium—Am 

36-3 Osmium—Os 


32-3 

Antimony—Sb 

36-4 Oxygen—O 


32-4 

Argon—A 

[5] 36-5 Palladium—Pd 

Hit 

32-5 

Arsenic—As 

36-6 Phosphorus—P 


32-6 

Astatine—At 

36-7 Platinum—Pt 


32-7 

Barium—Ba 

36-8 Plutonium—Pu 


32-8 

Beryllium—Be 

36-9 Polonium—Po 


32-9 

Bismuth—Bi 

36-x Potassium—K 


32-x 

Boron—B 

36-y Praseodymium—Pr 


32-y 

Bromine—Br 

(48) 37-0 Prometheium—Pm 

(44) 

33-0 

Cadmium—Cd 

37-1 Proactinium—Pa 


33-1 

Calcium-Ca 

37-2 Radium—Ra 


33-2 

Carbon—C 

37-3 Rhenium—Re 


33-3 

Cerium—Ce 

37-4 Rhodium—Rh 


33-4 

Cesium—Cs 

[6] 37-5 Rubidium—Pb 

[2] 

33-5 

Chlorine—Cl 

37-6 Ruthenium—Ru 


33-6 

Chromium—Cr 

37-7 Samarium—Sm 


33-7 

Cobalt—Co 

37-8 Scandium—Sc 


33-8 

Columbium—Cb 

37-9 Selenium—Se 


33-9 

Copper—Cu 

37-x Silicon—Si 


33-x 

Curium—Cm 

37-y Silver—Ag 


33-y 

Dysprosium—Dy 

(49) 38-0 Sodium—Na 

(45) 

34-0 

Erbium—Er 

38-1 Strontium—Sr 


34-1 

Europium—Eu 

38-2 Sulfur—S 


34-2 

Fluorine—F 

38-3 Tantalum—Ta 


34-3 

Francium—Fr 

38-4 Technetium—Tn 


34-4 

Gadolinium—Gd 

[7] 38-5 Tellurium—Te 

[3] 

34-5 

Gallium—Ga 

38-6 Terbium—Tb 


34-6 

Germanium—Ge 

38-7 Thallium—T1 


34-7 

Gold—Au 

38-8 Thorium—Th 


34-8 

Hafnium—Hf 

38-9 Thulium—Tm 


34-9 

Holmium—Ho 

38-x Tin—Sn 


34-x 

Hydrogen—H 

38-y Titanium—Ti 


34-y 

Indium—In 

(50) 39-0 Tungsten—W 

(46) 

35-0 

Iodine—I 

39-1 Uranium—U 


35-1 

Iridium—Ir 

39-2 Vanadium—V 


35-2 

Iron—Fe 

39-3 Ytterbium—Yb 


35-3 

Lanthanum—La 

39-4 Yttrium—Yt 


35-4 

Lead—Pb 

[8] 39-5 Zinc—Zn 

(41 

35-5 

Lithium—Li 

39-6 Zirconium—Zr 


35-6 

Lutecium—Lu 

39-7 


35-7 

Magnesium—Mg 

39-8 


35-8 

Manganese—Mn 

39-9 


35-9 

Mercury—Hg 

39-x 


35-x 

Molybdenum—Mo 

39-y Inorganic 


35-y 

Neodymium—Nd 

a Also: Helium—He; Krypton—Kr; 

(47) 

36-0 

Neptunium—Np 

Neon—Ne; Radon—Rn; Xenon—Xe 

• 

Numbers in parenthesis refer to columns used on the x-ray diffraction cards. 

t Numbers in brackets refer to columns used on the formula-name cards. 




203 



204 


PUNCHED CARDS 


Table 9-6. Radicals 


(51) * 40-0 aluminate 

40-1 ammonium 

40-2 antimonate 

40-3 antimonite 

40-4 arsenate 

40-5 arsenide 

40-6 arsenite 

40-7 bismuthate 

40-8 borate 

40-9 boride 

40-x bromate 

40- y bromide 

(52) 41-0 carbamate 

41- 1 carbide 

41-2 carbonate 

41-3 cerate 

41-4 chlorate 

41-5 chloride 

41-6 chlorite 

41-7 chromate 

41-8 cyanamid 

41-9 cyanate 

41-x cyanide 

41- y ferrate 

(53) 42-0 ferrite 

42- 1 fluoride 

42-2 fulminate 

42-3 germanate 

42-4 hafniate 

42-5 hexammine 

42-6 hydride 

42-7 hydroxide 

42-8 iodate 

42-9 iodide 

42-x manganate 

42- y molybdate 

(54) 43-0 nitrate 

43- 1 nitride 

43-2 nitrite 

43-3 osmate 

43-4 oxide 

43-5 pentammine 


43-6 phosphate 

43-7 phosphide 

43-8 phosphite 

43-9 plumbate 

43-x plumbide 

43- y rhenate 

(55) 44-0 selanate 

44- 1 selenide 

44-2 selenite 

44-3 silicate 

44-4 silicide 

44-5 stannate 

44-6 stannide 

44-7 sulphate 

44-8 sulphide 

44-9 sulphite 

44-x tantalate 

44- y telluride 

(56) 45-0 tellurite 

45- 1 thionate 

45-2 titanate 

45-3 thorate 

45-4 tungstate 

45-5 uranate 

45-6 vanadate 

45-7 zincate 

45-8 zirconate 

45-9 zirconyl 

45-x platinate 

45- y platinite 

(57) 46-0 chromite 

46- 1 gallate 

46-2 palladite 

46-3 

46-4 

46-5 

46-6 

46-7 

46-8 

46-9 

46-x 

46-y 


* Numbers in parenthesis refer to columns used on x-ray diffraction cards. 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 205 


information. The use of suffixes and prefixes to further qualify these 
radicals has not been attempted; one will find, for example, that all phos¬ 
phates, whether pyro-, ortho-, meta-, etc., will be coded by a punch at 6 
in column 43. (See Tables 9-5 and 9-6). Finally, column 56 is reserved for 
miscellaneous items of the inorganic code according to the following: 

56-0 Solid 
56-1 Liquid 
56-2 Gas 
56-3 Solution 
56-4 KBr Plate 
56-5 Hydrate 
56-6 Isotopic 

(3) Semi-empirical Formula. Columns 58 through 62 provide space to 
record the number of C, N, O and S atoms in the compound being indexed. 
These values are punched directly into the columns as labeled on the card. 
Provision for indicating a larger number of atoms than 9 in columns 60, 61 
and 62 is achieved by the use of overpunches. A “y” overpunch adds 10 to 
the value punched into the column, an “x” adds 20 and a “0” adds 30 to 
the number. A “0” punch alone would mean 30 atoms. 

(4) Melting or Boiling Point. A melting or boiling point is punched into 
columns 63 through 65. Melting points are used when the material is a solid 
at and above 20°C and boiling points are at or near 760 mm pressure only. 
The following code identifies the number punched into these columns: 

65-y—boiling point above 0°C 

64- y—boiling point below 0°C 

65- x—melting point. 

(5) Columns Reserved by AJS.T.M.. Columns 29 through 31 have been 
set aside for possible future use as determined by A.S.T.M. Committee 
E-13 and should not be used for private codes because of possible future 
conflicts. 

(6) Columns Reserved for Private Use. Columns 66 through 70 may be 
used by individuals as they see fit. At no time will Committee E-13 make 
code assignments to these columns. Such a section for private use will be 
found on all cards issued by the Committee. 

(7) Reference or File Number. Columns 71 through 80 are used to describe 
the location of the standard or original spectral data from which the card 
was prepared. Since it is common practice to publish a number of infrared 
spectra on the same page in journals, it is not possible to refer to a specific 
spectrogram by means of an ordinary journal reference. Therefore, spectra 
abstracted by A.S.T.M.-sponsored groups are assigned serial numbers and 
a numerical index of these serial numbers together with the name of the 



200 


PUNCHED CARDS 


compound and the journal reference is published for users of the cards. 
Regular catalogs of infrared spectra bear individual serial numbers in which 
case the same serial numbers are used on the cards. 

Letters punched into column 79 differentiate between the several collec¬ 
tions of spectra according to the following code: 

79-A—American Petroleum Institute 
79-B—Users own file 
79-C—Sadtler Catalogue of Spectra 
79-D—NRC-NBS File of Spectra 
79-E—Spectra abstracted by A.S.T.M. 

79-F—Documentation of Molecular Spectroscopy File. 

Column 80 codes the type of data carried by the card and the following 
assignments have been made: 

80-A—Infrared Absorption Data 
80-B—X-Ray Diffraction Powder Data 
80-C—Ultraviolet Absorption Data 
80-D—Visible Absorption Data 
80-E—Mass Spectral Data 
80-F—Raman Data 
80-G—Subject-Author Data 
80-H—Near Infrared Data 

Other assignments will be made as additional types of data are handled in 
the cards. 

Workers at Dow Chemical Company 16 have developed a method of identi¬ 
fying the components of a mixture by means of the infrared spectrogram, 
IBM cards and a collator. The stronger bands of standard spectra are 
coded by 9-row punches into columns representing the spectrum from 5 to 
16 microns at intervals of from 0.1 to 0.5 microns. A selected list of struc¬ 
ture groupings and classes are coded by “x” overpunches into these same 
columns. A collator can be wired to segregate in a single pass of the cards 
those representing spectra that (1) do not have bands at arbitrarily desig¬ 
nated wavelength positions, (2) do have bands at other designated positions 
and (3) do have any number of designated grouping or class codes. This 
reduces the sorting time for the identification of mixtures. It was neces¬ 
sary to construct a special plugboard to simplify the rather time-consuming 
operation of wiring the standard IBM plugboard for each run. 

Investigations carried on at Tennessee Eastman Company 1 * have shown 

15 Baker, A. W., N. Wright, and A. Opler, “Automatic Infrared Punched-Card 
Identification of Mixtures,” Analytical Chemistry, 25, 1457-60, (1953). 

“Otis, M. V., “A Statistical Study of the Wyandotte-A.S.T.M. Punched Card 
Library of Infrared Absorption Spectra,” presented at the Pittsburgh Conference on 
Analytical Chemistry and Applied Spectroscopy, Pittsburgh, Pennsylvania, March 
3, 1955. 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 207 


that the Wyandotte-A.S.T.M. cards can be handled in much the same way 
on the IBM Electronic Statistical Machine (101) so that a number of sorts, 
both positive and negative in any area of the card, can be made simul¬ 
taneously with one paas of the cards through the machine to reduce sorting 
time. A unique feature of this method is the segregating of cards in several 
pockets depending upon whether they represent possible components of 
binary, ternary, quaternary or higher mixtures. 

Near Infrared Absorption Spectroscopy 

The recent development of specialized equipment has permitted a greatly 
increased use of the near infrared region of the spectrum for analytical 
purposes. Attempts to incorporate these new data into the existing code 
systems for the regular infrared region were generally unsatisfactory because 
a closer punching resolution for band positions is desirable. Altho a special 
hand-sorted card has not been proposed as yet, the Standard Data Sub¬ 
committee of ASTM Committee E-13 has approved an IBM card system 
to complement its existing systems and make better use of these new near 
infrared data. Since the purpose of the near infrared cards is essentially 
the same as that of the regular infrared cards previously described, the 
same codes apply to both in all respects except the coding of absorption 
data. Therefore, only those codes that apply to the punching of the spectral 
data are discussed here. 

The near infrared spectral data indexing system for IBM cards is designed 
to handle absorption data in the region from 0.70 through 3.59 microns with 
a punching resolution of 0.01 micron. To be consistent with the other 
Wyandotte-ASTM Cards, band positions are coded in terms of wavelength 
in microns. At the head of columns 1 through 29 on the IBM card are 
printed numbers from 0.7 through 3.5 at intervals of 0.1 for each column. 
These numbers represent the whole number and tens values of the band 
positions and the hundreds values are punched into appropriate columns. 
Thus, the number printed at the head of column 20 is “2.6” so that a band 
position of 2.65 microns is coded by a single punch in column 20 at the “5” 
position. Determination of which bands to code follows the same general 
rules as prescribed for coding regular infrared spectra and sorting operations 
follow the same procedures previously described. It will be noted that the 
letter “H” in column 80 denotes near infrared data. 

X-Ray Diffraction Powder Analysis 

Qualitative analytical chemistry making use of x-ray diffraction powder 
data, essentially involves matching the diffraction pattern obtained from 
an unknown crystalline material with an identical pattern in a file of stand¬ 
ard data obtained from known materials. Such a pattern consists of a num¬ 
ber of diffraction lines of varying intensity separated by varying distances. 



208 


PUNCHED CARDS 


Table 9-7. Hanawalt Groups Number Code 


d values 

Group do. 

d values 

Group 

Under 0.80 

1 

3.05-3.09 

45 

0.80-0.89 

2 

3.10-3.14 

46 

0.90-0.99 

3 

3.15-3.19 

47 

1.00-1.04 

4 

3.20-3.24 

48 

1.05-1.09 

5 

3.25-3.29 

49 

1.10 1.14 

6 

3.30-3.34 

50 

1.15-1.19 

7 

3.35-3.39 

51 

1.20-1.24 

8 

3.40-3.44 

52 

1.25-1.29 

9 

3.45-3.49 

53 

1.30-1.34 

10 

3.50-3.59 

54 

1.35-1.39 

11 

3.60-3.69 

55 

1.40-1.44 

12 

3.70-3.79 

56 

1.45-1.49 

13 

3.80-3.89 

57 

1.50-1.54 

14 

3.90-3.99 

58 

1.55-1.59 

15 

4.00-4.09 

59 

1.60-1.64 

16 

4.10-4.19 

60 

1.65-1.69 

17 

4.20-4.29 

61 

1.70-1.74 

18 

4.30-4.39 

62 

1.75-1.79 

19 

4.40-4.49 

63 

1.80-1.84 

20 

4.50-4.59 

64 

1.85-1.89 

21 

4.60-4.69 

65 

1.90-1.94 

22 

4.70-4.79 

66 

1.95-1.99 

23 

4.80-4.89 

67 

2.00-2.04 

24 

4.90-4.99 

68 

2.05-2.09 

25 

5.00-5.24 

69 

2.10-2.14 

26 

5.25-5.49 

70 

2.15-2.19 

27 

5.50-5.74 

71 

2.20-2.24 

28 

5.75-5.99 

72 

2.25-2.29 

29 

6.00-6.49 

73 

2.30-2.34 

30 

6.50-6.99 

74 

2.35-2.39 

31 

7.00-7.49 

75 

2.40-2.44 

32 

7.50-7.99 

76 

2.45-2.49 

33 

8.00-8.49 

77 

2.50-2.54 

34 

8.50-8.99 

78 

2.55-2.59 

35 

9.00-9.49 

79 

2.60-2.64 

36 

9.50-9.99 

80 

2.65-2.69 

37 

10.0-10.9 

81 

2.70-2.74 

38 

11.0-11.9 

82 

2.75-2.79 

39 

12.0-13.9 

83 

2.80-2.84 

40 

14.0-15.9 

84 

2.85-2.89 

41 

16.0-17.9 

85 

2.90-2.94 

42 

18.0-19.9 

86 

2.95-2.99 

43 

20.0-and over 

87 

3.00-3.04 

44 




A.S.T.M. X-Ray Diffraction Data Cards. 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 209 


B 

B 


1 


ra 






















E 

a 





* FIRST LINK 

I#! 0;0 O 

'4 o'o o 

GROUP COOf 

? ? 
O 0*0 o 

_Z _ i _ £ _ L. 

O 
♦ *o 

TtHJ 

.9. 

fl 

D 

*.47 


4.09 

vu 

100 


70 

100 


Ou*| O 
4 ** I 

•j • 

Os;#; O 
_2_£_!_ 


UNI 

o ;#£o o 

/ ,4* * » 

9 ^59 9 


♦ 3i 

•t 


2-i 


Alph* Silicon Dl 0BU4# 


•OiO i 


Alpha CrtttobaUU 


RaA A 1.790* ftlur fi,J, 

Ola. 149.2 m CM #4 Coll 

1/1. VI# Ml dcorr aba* 

RM Clark, J. *». Cor. Soe., *9, 2$ (IMS) 


Syo Orthorhooblc &.G t* Pf 1 2 1 * t 

a. 7.00 A. 7.00 c. 7.00 A C 

• # T 19 

Ral In 


I# ■«# »T 9»fn 

*V D I.M* mp Color Co lor loss 

** C.C. 

lot flrlngt Roatod 4 hr*, to 1400"C., hold 
fin hour#. 

2nd flrlnfi Haatod S hr#, to 1400*C., hold 
tharo 24 tour#. 

iyi call# n-crtatobaUlo tho loo i«p. for*. 
Proa firIn* data, oatorial 1# proaxaoblr high 


71 

■231 

■ ^1 

~7T~' 

msm\ 

4.09 

100 


1.99 

10 

9.13 

60 


1.56 

IO 

2.94 

K> 


1.35 


2.47 

80 


1.94 


2.45 

10 


1.33 


2.10 

20 


1.29 


2.01 

20 


1.28 


1.92 

40 


1.23 


1.86 

40 


1.22 

10 

1.7$ 

10 


1.20 

10 

1.72 

10 


1.18 

10 

1.68 

20 


1.17 

10 

1.69 

10 


1.16 

10 

1.60 

40 


1.10 

10 


10 


1.09 

20 

1.56 

io i 

1## #l#o 

D- 9.99 

3.19 2. 

1.S2 

20 


4.90 

4.08 3, 

1.49 

20 


3.9$ 

1.613 

1.49 

20 





OOOOOOftOO 

I z ff 4 • 9 \l\ 9 C lo»! 

o o ♦. o o o m\ o 

_LJ_ 


S5TI 


»o 


^ o 

40 40 

-O -O* 


>jo 

ko *o* 

c-x-«*« ■« 

0>O nQ 

-o 


siOgil 


io o* 


o» 


• I Tl 

o o 

II Tl 

o o 


' * * e • i 

O O O O O O O 

> t s « • l 

O O O O O O O 


o 

o o 


Figure 9-4. A.S.T.M. Keysort X-Ray Diffraction Card. 


The pattern may be recorded on photographic film or on a chart by means 
of a recording potentiometer. At present the American Society for Testing 
Materials publishes pattern data for over 5,000 compounds in terms of sets 
of line spacings and densities. This mass of data offered an excellent oppor¬ 
tunity for the application of punched card methods. 

Hand-Sorted Cards. An early application of hand-sorted methods to 
the problem of handling x-ray diffraction data was made at Canadian In¬ 
dustries Limited, McMasterville, Quebec. 17 ' 18 These cards are now pub¬ 
lished and distributed by the American Society for Testing Materials, and 
a pamphlet describing the notching and sorting operations is available. 19 
The system makes use of the spacing values of the three strongest lines of 
the powder pattern and the identity of the elements in the compounds in¬ 
volved to provide notches for sorting. The Hanawalt method 12 of employing 
three cards for each compound is used wherein each of the three strong 
lines forms the basis of assigning a card to a group. The group number as 
determined by the Hanawalt Groups Code (Table 9-7), is punched into the 
group code field of the card (Figure 9-4) and the other two lines are punched 

17 Matthews, F. W., “Punched-Card Code for X-Ray Diffraction Powder Data,” 
Analytical Chemistry, 21, 1172-75 (1949). 

,s Matthews, F. W., “Tabulation of X-Ray Diffraction Powder Data for Chemical 
Analysis,” Can. Chem. and Process Ind., 31, 63-4 , 67-8, 71 (1947). 

19 “Instructions on Notching and Sorting Keysort X-Ray Diffraction Data Cards,” 
the American Society for Testing Materials, Philadelphia, Pennsylvania. 






































210 


PUNCHED CARDS 


Table 9-8. Code for Chemical Composition 


Aluminum 

A1 

3-2 

Americium 

Am 

13-12 

Antimony 

Sb 

9-7 

Arsenic 

As 

9-6 

Barium 

Ba 

2-12 

Beryllium 

Be 

2-8 

Bismuth 

Bi 

9-8 

Boron 

B 

3-1 

Bromine 

Br 

11-4 

Cadmium 

Cd 

4-10 

Calcium 

Ca 

2-10 

Carbon 

C 

7-2 

Cerium 

Ce 

12-7 

Cesium 

Cs 

1-6 

Chlorine 

Cl 

11-3 

Chromium 

Cr 

8-9 

Cobalt 

Co 

5-3 

Columbiuin 

Cb 

8-11 

Copper 

Cu 

5-6 

Curium 

Cm 

13-1 

Dysprosium 

t>y 

12-2 

Erbium 

Er 

12-3 

Europium 

Eu 

12-13 

Fluorine 

F 

11-2 

Gadolinium 

Gd 

12-1 

Gallium 

Ga 

3-6 

Germanium 

Ge 

4-12 

Gold 

Au 

5-11 

Hafnium 

Hf 

7-6 

Holmium 

Ho 

12-2 

Illinium 

11 

12-10 

Indium 

In 

3-7 

Iodine 

I 

11-5 

Iridium 

Ir 

6-13 

Iron 

Fe 

5-2 

Lanthanum 

La 

12-6 

Lead 

Pb 

4-1 

Lithium 

Li 

1-2 

Lutecium 

Lu 

12-5 

Magnesium 

Mg 

2-9 

Manganese 

Mn 

8-10 

Mercury 

Hg 

4-11 

Molybdenum 

Mo 

8-12 


Neodymium 

Nd 

12-9 

Neptunium 

Np 

13-10 

Nickel 

Ni 

5-4 

Nitrogen 

N 

9-4 

Osmium 

Os 

6-12 

Oxygen 

O 

10-9 

Palladium 

Pd 

6-11 

Phosphorus 

P 

9-5 

Platinum 

Pt 

6-1 

Plutonium 

Pu 

13-11 

Polonium 

Po 

10-1 

Potassium 

K 

1-4 

Praseodymium 

Pr 

12-8 

Protactinium 

Pa 

13-8 

Radium 

Ra 

2-13 

Rhenium 

Re 

8-3 

Rhodium 

Rh 

6-10 

Rubidium 

Rb 

1-5 

Ruthenium 

llu 

6-9 

Samarium 

Sin 

12-11 

Scandium 

Sc 

3-4 

Selenium 

Se 

10-12 

Silicon 

Si 

7-3 

Silver 

Ag 

5-7 

Sodium 

Na 

1-3 

Strontium 

Sr 

2-11 

Sulfur 

S 

10-11 

Tantalum 

Ta 

8-1 

Tellurium 

Te 

10-13 

Terbium 

Tb 

12-1 

Thallium 

T1 

3-8 

Thorium 

Th 

13-7 

Thulium 

Tm 

12-3 

Tin 

Sn 

4-13 

Titanium 

Ti 

8-13 

Tungsten 

W 

8-2 

Uranium 

U 

13-9 

Vanadium 

V 

8-7 

Ytterbium 

Yb 

12-4 

Yttrium 

Y 

3-5 

Zinc 

Zn 

4-9 

Zirconium 

Zr 

7-5 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 211 

into the second and third line fields with the 7-4-2-1 code. This system 
makes it possible to use any one of the three strongest lines to determine a 
Han await Group in which the sought-for card must be, and then to sort 
notches on the second and third lines to Isolate the desired cards. The cards 
are filed according to the groups so that the first sort results from the act 
of removing the cards from the file. A code designation of the elements in 
the compound (Table 9-8) is notched into the chemical composition field 
of the card and may be used as a cross-sort at any time such information is 
known. Considerable space is left in the edges of the card for notching ad¬ 
ditional information such as melting point, indices of refraction, etc., but 
this can be done at the discretion of each individual user. On the face of the 
card are printed the complete crystallographic and x-ray powder data of 
the compound together with name, formula, purity, source and other perti¬ 
nent notes. A system for tabulating detailed crystallographic data in Key- 
sort cards has been proposed by workers at Armour Research Foundation. 10 
These cards contain all of the data normally published in the Crystallo¬ 
graphic Data series in “Analytical Chemistry” and provides an excellent 
method of searching the data for identification purposes. 

Machine-Sorted (IBM) Cards. A method of applying International 
Business Machines Company cards and equipment to the problems of 
sorting and correlating x-ray diffraction powder data was developed at 
Wyandotte Chemicals Corporation. 21 Through the cooperation of members 
of the Joint Committee on Chemical Analysis by Powder Diffraction Meth¬ 
ods this system was accepted by the A.S.T.M. and punched cards indexing 
all of the powder diffraction data are available. Supplements of the IBM 
cards are offered with each set of the x-ray data as they are released by 
A.S.T.M. The Wyandotte-A.S.T.M. x-ray diffraction card, codes and sort¬ 
ing methods are modifications of the infrared absorption indexing system 
previously described and much of the descriptive material pertaining to the 
handling of IBM cards presented in the infrared section is equally appli¬ 
cable here. The following description includes the necessary codes that 
enable one to prepare and use the cards. Additional details and examples 
may be obtained from the instruction booklet distributed by A.S.T.M. 21 

The card for indexing x-ray diffraction data (Figure 9-5) is divided into 

*®McCrone, W. C., “Punched-Card System for Tabulating of Crystallographic 
Data,” Analytical Chemistry , 28, 972-5 (1956). 

*' Kuentzel, L. E., “The Use of Hollerith Punched Cards for Indexing X-Ray Dif¬ 
fraction Powder Data,” presented before the American Crystallographic Associa¬ 
tion, Chicago, Illinois, October 24, 1951. 

** Kuentzel, L. E., “Codes and Instructions for Wyandotte-A.S.T.M. Punched 
Cards Indexing X-Ray Diffraction Powder Data,” Wyandotte Chemicals Corpora¬ 
tion, Wyandotte, Michigan (1951) and IBM Technical Newsletter No. 4, Interna¬ 
tional Business Machines Corporation, New York City (1953). 



212 


PUNCHED CARDS 


»-•*» •trr*Mri«a 



MNI t» t* 

21222122222222222222221222%1 12121 
■ •Uf OittlKtiO* MU 

ummminmnnmn^nmm 

C»«» till 

44444444444444444444444444^444444444 

*W»l'«MCO »r 

4 S 5 5 5 5 S 5 S 5 5 5 5 5 S 5 5 5 5 S 5 5 5 5 5 5 5 S 5 J 5 5 i\ 

f«t tociff* tt* TUT INC ■•▼*•)*(.« 

Mill 18(188 II I 8 I 8 8 8 8 8 8(1(166(11(8 

Itio (M( »T«||T 

111111111111111111111111111111111] 
MUItliTml 0 . M 

i i i a i i i i i i i i i i a i i i i i a i i i i lit i • i i i i 




mnmmmmmnnmmn 

i • t i i i •attitHMawi’iaioanani'SKnaiiBx 
■a tMMi cor«**«T '••! at KtTOOTH MKUil coaaoa*Ti««. tuiMiTi, •<« 


722 


n 


iiiiiiiiiii 1 

u •••«>•• wt< » u 

I I I 1 I 1 1 I 1 1 II f 


22222222 


22 


snnnmm 

j 


44 4 
S S 5 


4444444444 


>SSSSS»sjs* 

i a j a 111111 (11 

! I* 

i i i\i i n in hi i 

aaiHiaaiaaia 


rai iTiaTataara 1 


aaaiaTfi 

a ■ a»»a a •» m a a *> 

1 I 1 1 1 1 1 I 111 I ill 1 1 I 111 111 

2 2 2 2 2 2 7 2 11 2 2l2 2 2 2 2*2 2*2 2 

tm« laaci'aa t*«T(a 

mnnmnnnnm 

•ot ai«fi.o*ta « 

444444444444444444444 

tniiici actcaaca moMTam 

nnmsmminmi 

NIIITCa ta NKldMiT Ok 

(8(11 I 1 1 1 8 ( 1 8 8 6 1 8 til I 

aT*aaoTTt (»i«iuii caio««t«« 
111111111111111111111 
BT*«e«TTf. «<«(«*■ 

iiiii|i mill i ii 11 ii ii 

i ! ; 

nmmmmnmn 



Figure 9-5. Wyandotte-A.S.T.M. X-Ray Diffraction Data Card. 


the following areas for coding purposes: 

(1) Diffraction Line Spacings—columns 1 through 35 

(2) Hanawalt Group Code—columns 36 and 37 

(3) Chemical Classification—columns 43 through 62 

(4) Melting Point—columns 63 through 65 

(5) Reserved for A.S.T.M.—columns 38 through 42 

(6) Reserved for Private Use—columns 66 through 70 

(7) Reference or Serial Number—columns 71 through 80 

All available data about a given compound are coded into one card and 
three copies of the card, identical except for the group code number, are 
included in the file. Punches at 1, 2 or 3 in column 27 indicate which of the 
three strongest lines was used in determining the group code for the particu¬ 
lar card. 

(1) Diffraction Line Spacings: All “d” lines of the diffraction pattern 
having an intensity one-tenth or more of the intensity of the strongest line 
are punched into the cards. This is achieved by punching the final digit into 
the column headed by the number supplying the rest of the digits. The col¬ 
umn headings are printed on the card but Table 9-9 gives the column head¬ 
ing code. 

Values below 1.00 A are rounded off to the nearest tenth angstrom and 
punched directly into column 1. From 1.00 through 3.40 angstrom units 
the punching resolution is 0.01 unit and the value of the hundreds digit is 
punched into the column having the proper heading. Thus, a value of 1.98 A 
is coded by a single punch at the 8 position in column 11 which is headed 
by the number 1.9. Beginning with column 27 and through 33 the punching 
resolution is again 0.1 A. Thus, a value of 23.7 A is coded as a 4 punch 
in column 35. All values greater than 29 A are coded as a 9 punch in column 
35. Use of the selector switch on the sorter, as described in the discussion 
on the infrared system, enables one to segregate all cards bearing any given 


















QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 213 


Table 9-9. Line Spacing Code 


Column 

Line 

Column 

Line 

1 

0 

19 

2.7 

2 

1.0 

20 

2.8 

3 

1.1 

21 

2.9 

4 

1.2 

22 

3.0 

5 

1.3 

23 

3.1 

6 

1.4 

24 

3.2 

7 

1.5 

25 

3.3 

8 

1.6 

26 

3.4 

9 

1.7 

27 

3 

10 

1.8 

28 

4 

11 

1.9 

29 

5 

12 

2.0 

30 

6 

13 

2.1 

21 

7 

14 

2.2 

32 

8 

15 

2.3 

33 

9 

16 

2.4 

34 

10 

17 

2.5 

35 

20 

18 

2.6 




line combination. Also, as with the infrared system, one can sort over a 
narrow range of values simultaneously if the exact value of the line is in 
doubt. With these cards, since a great many lines are coded in each, sorting 
operations are not confined to the three strongest which may be less charac¬ 
teristic than other lines in the pattern. 

(2) HanawaU Group Code. Use is made of the three-card-per-compound 
and group code system for filing the punched cards in Hanawalt groups, 
as described previously for hand-sorted x-ray diffraction cards. This enables 
one to use any one of the three strongest lines to determine a group in which 
the wanted card must be and then to withdraw only that group from the 
files for sorting operations. For reason of uniformity, the same group code 
as suggested by Dr. Hanawalt and used on the Keysort cards previously 
described is used on the IBM cards (Table 9-7). The code group numbers 
are punched directly into columns 36 and 37. This enables the cards to be 
sorted into Hanawalt groups by machine for filing purposes. 

(3) Chemical Classification. The identity of all elements, inorganic radi¬ 
cals, type of compound and other pertinent data are coded into this section. 
This enables one to make card-eliminating sorts in this section when such 
information is available. This has proved to be a particularly powerful tool 
whenever it can be used. The codes involved are direct and the methods 
employed are similar to those described for structure sorts in the section 
on infrared. The Elements Code for section A on the card and the Radicals 
Code for section B are the same as used for inorganic compounds on the 
infrared card except that different columns are involved. The column 
values for use on the x-ray cards will be found in parentheses in Tables 9-5 



214 


PUNCHED CARDS 


and 9-6. Codes for sections C and D on the x-ray card are given below: 

Section C, Organic 
59-0 Saturated aliphatic 
59-1 Unsaturated aliphatic 
59-2 Saturated monocyclic 
59-3 Unsaturated monocylic 
59-4 Saturated polycyclic 
59-5 Unsaturated polycyclic 
59-6 Benzo aromatic 
59-7 Polybenzo aromatic 
59-8 Fused ring aromatic 
59-9 Heterocyclic 
59-x Unassigned 
59-y Unassigned 

Section D, Miscellaneous 
61-0 Hydrated 
61-1 Inorganic 
61-2 Organic 
61-3 Metal organic 
61-4 Unassigned 

(4) Melting Point. Columns 63, 64 and 65 provide for recording melting 
points. All melting points higher than 999°C are punched as 999. If the value 
punched into the card is a negative number, the fact is indicated by an ad¬ 
ditional “x” overpunch in column 65. Melting points have not been gener¬ 
ally available together with the diffraction data, so this feature of the cards 
as currently distributed is far from complete. 

(5) and (6) Reserved Areas. Columns 38 through 42 are reserved for 
future use by A.S.T.M., and columns 66 through 70 are reserved for private 
use by individual laboratories. 

(7) Reference or Serial Number. This section of the card provides space to 
record a reference to the original data from which the card was prepared. 
Provision is made for either a journal or book reference or to the A.S.T.M. 
data card serial number. So far, only the latter type of reference has been 
used on published punched cards. The serial number is punched directly 
into columns 75 through 78, the set designation is coded into column 79 
where “A” means Set 1, “B” means Set 2, “C” means Set 3, etc. On the 
A.S.T.M. x-ray data cards, the sets are designated by a digit in front of 
the serial number. Thus, data identified by serial number 3-0725 on the 
A.S.T.M. x-ray data card is punched into an IBM card bearing the number 
725C. The code used for punches into column 80 is the same for all types 
of Wyandotte-A.S.T.M. cards and the assignments are listed earlier in 
the infrared section of the chapter. 

Normally, the x-ray IBM cards are filed according to Hanawalt Group 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 215 


numbers so that any given group or groups may be removed directly from 
the files for sorting operations. Any one of the three strongest lines may be 
used to determine group selected. The first actual sort on the cards will 
depend upon what other information is available. If the pattern alone must 
form the basis of the search, the most uncommon line should be used for 
the first sort in order to eliminate as many cards as possible. This may be 
the innermost line or an unusually strong line in the short-spacing section 
of the pattern. The operations are carried out in exactly the same way as 
described for the sorting of infrared absorption bands earlier in this chapter. 
Additional sorts on each residue deck continue to eliminate unwanted 
cards. If a reliable sort can be made on a metal element, the search can be 
narrowed very rapidly. In any event, the order of the sorting operations 
does not alter the final results and because of the relatively small number 
of cards in the Hanawalt Groups, the operations require but a very few 
minutes. The serial number on the final card or cards refers one directly 
to the complete data and names in the A.S.T.M. powder diffraction file 
for a final comparison with the unknown before the identification is ac¬ 
cepted. A recent paper by Beukelman WA describes efficient sorting operations 
for effective use of these cards. 

Ultraviolet Absorption Spectroscopy 

The application of ultraviolet absorption data to qualitative analytical 
determinations involves much that is similar to the use of infrared absorp¬ 
tion data. The ultraviolet spectrogram is a fingerprint of the compound 
that produced it, although it is usually somewhat less detailed than the 
normal infrared spectrogram. In such matters as the coding and sorting of 
absorption band positions, the identity of important elements, a classifica¬ 
tion of the chemical structure and melting or boiling points, both methods 
are quite similar. Moreover, the fact that ultraviolet absorption data can be 
recorded and published in a number of different ways provides the same 
problems of matching unknown data with published data. However, in 
general, the ultraviolet spectra offer less detail for comparison purposes 
and the effects of solvents are more pronounced. This is reflected in the 
design of cards to index ultraviolet data for sorting purposes. It becomes 
accessary to include details on the intensity of critical absorption bands 
and to identify the solvent used. The very large mass of published ultra¬ 
violet spectral data has made the use of punched card indexing methods for 
universal searches a necessity. 

Hand-Sorted Cards. At present there is no widely accepted and used 
notched card system for handling ultraviolet absorption data. The Na- 

** A Beukelman, T. E., “Efficient Use of IBM File of ASTM Powder X-Ray Diffrac¬ 
tion Data,” Analytical Chemistry, 29, 1269-72 (1957). 



216 


PUNCHED CARDS 


tional Research Council Committee on Spectral Absorption Data 7 is 
working on Keysort card to be a companion to the infrared card currently 
being distributed. Present status of the card provides for coding principal 
absorption bands from 200 to 400 millimicrons, a field for band-no band 
coding in 50 millimicron steps, chemical structure, solvents and specific 
absorbence values of the strongest bands. Printed on the card will be the 
name and formulas of the compound, physical state, solvent concentration, 
cell thickness, source and purity of compound and contributing laboratory. 
Meanwhile, another proposal for such a card has been published** and 
general interest is rapidly increasing. However, since none of the cards or 
systems is being offered commercially at the present, the codes are not in 
sufficiently final form to be included here. 

Machine-Sorted (IBM) Cards. A logical extension of the methods used 
in handling infrared absorption data into the field of ultraviolet and 
visible asborption spectroscopy was made at Wyandotte Chemicals Cor¬ 
poration.* 4 This IBM system, with modifications contributed by members 
of A.S.T.M. Committee E-13, has been adopted by the American Society 
for Testing Materials as a standard method of indexing and sorting such 
data 14 and large decks of punched and printed cards are available from the 
Society. The collecting and abstracting of ultraviolet absorption data for 
these cards are in the hands of the same A.S.T.M. committees and the 
cards are being prepared by the same group at the National Bureau of 
Standards as described earlier in this chapter. The sorting techniques out¬ 
lined for use with the infrared Wyandotte-A.S.T.M. cards are applicable 
to the ultraviolet cards to be described. The following description includes 
all of the codes necessary for the proper use of the cards in searching pub¬ 
lished ultraviolet absorption data for qualitative analytical purposes. 

The Wyandotte-A.S.T.M. card indexing ultraviolet absorption data (see 
Figure 9-6) is divided into the following areas for coding purposes: 

(1) Ultraviolet Absorptions—columns 1 through 11 

(2) Number of Peaks—columns 12 and 13 

(3) Intensity of Peaks—columns 14 through 17 

(4) Solvents and pH—columns 29 through 31 

(5) Chemical Classification—columns 32 through 57 

(6) Semi-empirical Formula—columns 58 through 62 

(7) Melting or Boiling Point—columns 63 through 65 

(8) Reserved by A.S.T.M.—columns 18 through 28 

(9) Reserved for Private Use—columns 66 through 70 

(10) Reference or Serial Number—columns 71 through 80 

** Kendall, C. E., “Indexing of Data on Ultraviolet Ansorption Spectroscopy,” 
Applied Spectroscopy, 9, 158-165 (1955). 

14 Kuentzel, L. E., “The Indexing and Sorting on IBM Equipment of Infrared, 
Ultraviolet, Mass and Other Standard Data,” paper presented at the Pittsburgh 
Conference on Analytical Chemistry and Applied Spectroscopy, Pittsburgh, Pennsyl¬ 
vania, March 6, 1952. 




QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 217 



pimMirn 




i ijilililih*"n n i*"m n 

I lit If ill >f 22222222222 


)issit)i»]itaiiiiii« 




SIIBISIISSSSIISISIt Ijf 4jl 

cm MMTI rm tiitw utimii 


111111111111111111111 


M 


:: 


iiiiiili 


111111111111111111 
lilt 2tilt!!!!!!! Sill III2b 2 
SllfSISSlSIlJItllllSISSItlSIlSSSI 


C»« M iOH . OkM* 



MB ffpli *a a n 


«*■*• MCMRCO iM TMK 

|l|«»IIIIISI»ISIS»»SSISSimSIS9ll9l 


WMMCM A-0O 
111111111111111111111111111111111 
WWNOOTTI CW«MIC>k» COMO Wi 

[III9IIIIIBIIIIIIIIIIIIIIIIIIIIIII 

WVWCOTTL monmam 


TTill 


itr^ 1 


himm 


i;imh 
b 2 2 2 2b lb 2b 2 2 20 

SlSIllllSStllSiS 


I1IIIISSIIIIIIB 

111111111111111 

riOM 






Figure 9-6. Wyandotte-A.S.T.M. Ultraviolet Data Card. 


All of the data from one compound are punched into one card. Codes for 
items (5), (6), (7) and part of (10) are identical to those used for the same 
columns on the infrared card and since they are described in detail in an 
earlier section of this chapter, they will not be repeated here. 

(1) Ultraviolet Absorptions. The coding of the positions of ultraviolet 
absorption bands or peaks is done in terms of wavelength in millimicrons. 
The coding resolution is 2 m/i, that is, any number of peaks 2 m/i or more 
apart may be coded individually and peaks closer than this are coded as 
one. The wavelength interval covered by each column is printed at the 
head of the column. Thus, a peak value of 244 m/x would be coded by a 
single punch at the number 2 position in column 3 which is headed by the 
number 240. Each successive digit in the column represents an increment of 
2 m/i over the preceding one and the 0 punch value is that which is printed 
at the head of the column. Again, a value of 338 m/i is coded by a punch 
of 9 in column 7 which has the value 320 at its head. To indicate the range 
of the spectra data covered by the particular spectrogram being coded, an 
“x” overpunch is placed in each column for which no data are available. 
A “y” overpunch in the same column where the last measurements on a 
terminal absorption are recorded indicates that there is a possible band just 
outside the range covered by the published data. Finally, the general 
position of the longest wavelength band in the spectrum is indicated with 
an “x” overpunch in the same column with the punch designation of the 
peak position. Such an “x” overpunch need not be confused with the “no 
data” overpunches because in the latter case no other punches Appear in 
the same column. 

Sorting the peak positions with the ultraviolet indexing cards is essen¬ 
tially the same as the methods used in handling the cards indexing infrared 
spectral data. Both positive >ind negative sorting approaches are feasible. 
Use of the longest wavelength band “x” overpunch is an effective tool. 
Thus, if there is an unknown with its longest wavelength band at 314 mil- 


































218 


PUNCHED CARDS 


Column 

Punch 

Table 9-10 

Range (mp) 

Peaks 

12 

0 

200 to 250 

None 

12 

1 

200 to 250 

One peak 

12 

2 

200 to 250 

Two peaks 

12 

3 

200 to 250 

Three or more 

12 

5 

250 to 300 

None 

12 

6 

250 to 300 

One peak 

12 

7 

250 to 300 

Two peaks 

12 

8 

250 to 300 

Three or more 

13 

0 

300 to 350 

None 

13 

1 

300 to 350 

One peak 

13 

2 

300 to 350 

Two peaks 

13 

3 

300 to 350 

Three or more 

13 

5 

350 to 400 

None 

13 

6 

350 to 400 

One peaks 

13 

7 

350 to 400 

Two peaks 

13 

8 

350 to 400 

Three or more 


limicrons, sorts can be made to eliminate all spectra that do not have their 
longest band in the 300 to 318 mji range (column 6) by sorting on the “x” 
overpunch. 

(2) Number of Peaks. The number of peaks in each 50 millimicron inter¬ 
val of the spectrum are indicated by punches in columns 12 and 13, accord¬ 
ing to the given code in Table 9-10. 

Thus, a spectrogram which exhibits peaks at 225, 314, 322, 345 and 375 
millimicrons would be coded by punches at 12-1 and 13-3,6. A special 
punch at “y” in column 12 is used to indicate that the data coded into 
the ultraviolet absorption section of the card was published in tabular form. 
In such cases no “x” overpunches are used except to locate the position of 
the longest wavelength band. 

(3) Intensity of Peaks. The intensity, in terms of absorbence for a solu¬ 
tion of 1 gram per liter in a 1-cm cell (the absorptivity), of the strongest 
peak in each 50-mjx interval is punched into columns 14 through 17 by 
means of the code given in Table 9-11. These data are included so that one 
can make use of the differences in peak intensities to separate spectra with 
peaks located at the same wavelength positions but exhibiting differences in 
intensity. 

(4) Solvents and pH. Because of the influence of solvent upon the shape 
of ultraviolet spectra, it was deemed advisable to provide a method of 
segregating spectra obtained from compounds in different solutions. In 
Table 9-12 is a direct code for a number of solvents commonly used. The 
solvents are arranged in order of frequency of use as revealed by a study 
of a large number of published spectra. 

In order to provide a finer breakdown of water solutions based upon pH, 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 219 


Table 9-11 


Column 

Range (idm) 

Punch 

Intensity 

14 

300 to 250 

0 

0 to 

1 

15 

250 to 300 

1 

1 to 

3 

16 

300 to 350 

2 

3 to 

10 

17 

350 to 400 

3 

10 to 

30 



4 

30 to 

50 



5 

50 to 

75 



6 

75 to 

100 



7 

100 to 

200 



8 

200 to 

300 



9 

300 to 1000 



X 

Over 1000 




Table 9-12 

Column 

Punch 

Solvent 

29 

0 

Aliphatic hydrocarbon; isooctane, cyclohexane, etc 

29 

1 

95% Ethanol 

29 

2 

Absolute ethanol 

29 

3 

Absolute methanol 

29 

4 

1 Normal or stronger acid 

29 

5 

0.1 Normal acid solution (HC1) 

29 

6 

Water; pH 5 to pH 9 

29 

7 

0.1 Normal base solution (NaOH) 

29 

8 

1 Normal or stronger base 

29 

9 

Dioxane and water mixtures 

29 

X 

CHClj , CC1 4 , SC 2 

29 

y 

Aromatic hydrocarbons; benzene, toluene, etc. 

30 

0 

Glacial acetic, cone. H 2 S0 4 , etc. 

30 

l 

Ethers 

30 

2 

Ketones, esters 

30 

3 

Pyridine and other basic solvents 

30 

4 

Dimethyl formamide, dimethyl acetamide 

30 

5 

Other 

30 

6 

Solvent unknown, not reported, etc. 

30 

7 

No solvent—vapor, film, liquid, gas, etc. 


a direct code Is supplied. This information, when available, is punched into 
column 31 according to the following code: 


Punch 

pH 

Punch 

pH 

0 

below 1 

5 

7 

1 

1 or 2 

6 

8 

2 

3 or 4 

7 

9 

3 

5 

8 

10 or 11 

4 

6 

9 

above 11 


(8) and (9). Reserved Sections. Columns 18 through 28 are reserved for 
future use by A.S.T.M. action, and columns 66 through 70 are reserved for 
private use as on the infrared indexing card. 



220 


PUNCHED CARDS 


(10) Reference or File Number. This section of the card is used to identify 
the source of the data coded into the card. As with the infrared indexing 
cards, it has not been feasible to use journal references directly so serial 
numbers are assigned and lists of serial numbers, compound names and 
journal references are issued by A.S.T.M. to users of the cards. Where 
publishers have already assigned serial numbers to the spectra, as is done 
by the American Petroleum Institute Research Project 44, the same serial 
numbers are used on the cards. Following are the code assignments to 
column 79 for indicating the source of ultraviolet data: 

Code Source 

A American Petroleum Institute Research Project 44 

B User’s own file of spectra 

C Spectra issued by the NRC-NBS Committee 

D Spectra abstracted by A.S.T.M.-sponsored groups 

Others will be added as needed. The code used in Column 80 is the same 
as that used on the infrared card and may be obtained by referring to the 
appropriate section of this chapter. 

Visible Absorption Spectroscopy 

The application of visible absorption data to qualitative analytical 
determinations follows essentially the same pattern as developed for ultra¬ 
violet absorption data. There is only the added feature that the compounds 
and solutions involved usually have color, and it is convenient to have a 
means of indicating this fact. Because of the similarity of the systems in¬ 
volving infrared, ultraviolet and visible absorption data for qualitative 
work, a general discussion will not be given here and the reader is referred 
to the appropriate previous discussion in this chapter for background in¬ 
formation. 

Hand-Sorted Cards. At present, there is no generally accepted and used 
punched card system for handling visible absorption data. The National 
Research Council Committee on Spectral Absorption Data 7 has plans for 
a Keysort card covering the range of 400 to 800 mji. It will be quite similar 
to the one nearly completed for the ultraviolet region which was previously 
described. The only major publication on the subject 2 * combines ultraviolet 
and visible data in one card at some sacrifice in coding resolution. 

Machine-Sorted (IBM) Cards. A system for handling visible absorp¬ 
tion data in IBM cards was proposed by workers at Wyandotte Chemicals 
Corporation. 24 This system, with modifications contributed by members 
of A.S.T.M. Committee E-13, has been adopted by the American Society 
for Testing Materials as a standard method of indexing and sorting such 
data 14 , and decks of punched and printed cards are available from the 
Society. The collection and editing of the published data and the prepara- 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 221 



mMil tncfUi MteartiM lift 

nssnmnisstssssssnji sjsl 


> > nift 11 Mirjiji 111 

11 ll 


lilt 


2 2 2 2 2l2 2 2 2 2 


44444444444444444444441| 

niniMTti 

ISSIISSSSSSS»$ISSS»!SS>| 

•km t«a«Tv ro« MiriN urcaiM 

• •1111111119111111411111 

im m« imir 

7 ? 7 n 7 ? m ? ? 7 7 j? ? 7 m n| 

•••••••••••••••••••••ll 

• iitniinii tin mini inn 

i i i • i • MiingaNaanMBiiaaiiiiii 



CMC MIC Aw CLAtairiCATlON 

L-l'.m 

icti 

Hi 

1 

1MC CMC (MIT , 

H 


4«4 


•M| 

77 

• • 


M 


1111 
2 2 2 2 

nil 

4444 


|9 S]»|I99I 

• til 
7 7 7 7 

• ••• 


• III 


WW»I»I HM •* ■VMMTTI 


•••(••••••••••••••h~m 

im MMMMMMM •»!•( tilt MM Will ■ 
111111111111111111 



222(222222222 22222222212222 2Q 22122 


13 ISin 1 2 1 1 2 ) 1 ) 1 2 1 2 2 2 2 1 1 1 nil 11 

▼ Ml* inou 

4444444444444444444444444444444 

WA* MMCMAMCO M Tuf 

Dmmimimiisssfsisssssii 

aMvtica mcmamcm i«ao*«Toa*a 

•••••fltSlftl« 8 BISft»ftS 9 IIISI§CCI 

RttCADCM AMO DCvCkOMCNT 0<Vi**0M 

7777777777777777777777777777777 

WVANOOTTt CMCMiCAlt C0*»0* ATOM 

••llltllllllllllllllllll•ll•tl• 

W V A MOOT T C , MiCmiOAM 

iiiitiiiiiiiiiini ii i inn inn! 




•ni.a 


2(2 21) 22 29 

1)2)1)333 

444444441 
SS9SMIIS 
• ••••mi 

777777777 

••••••••I 

!••••••••• 

i*«Miaaii*idb 


Figure 9-7. Wyandotte-A.S.T.M. Visible Data Card. 


tion of the cards are being handled by the same groups associated with the 
other Wyandotte-A.S.T.M. Cards as previously described. The codes and 
sorting instructions for use with these cards are very similar to those used 
on the ultraviolet cards. All codes necessary for the proper use of the cards 
are supplied herewith. 

The Wyandotte-A.S.T.M. card indexing visible absorption data (see 
Figure 9-7 is divided into the following areas for coding purposes: 

(1) Visible Absorptions—columns 1 through 10 

(2) Number of Peaks—columns 11, 12 and 13 

(3) Intensity of Peaks—columns 14 through 18 

(4) Color Index Number—columns 19 through 23 

(5) Solvents and pH—columns 20, 30 and 31 

(6) Chemical Classification—columns 32 through 57 

(7) Semi-empirical Formula—columns 58 through 62 

(8) Melting or Boiling Point—columns 63 through 65 

(9) Reserved by A.S.T.M.—columns 24 through 28 

(10) Reserved for Private Use—columns 66 through 70 

(11) Reference or Serial Number—columns 71 through 80 

All the data from one compound are coded into one card. Codes for items 
(5), (6), (7), (8) and most of (11) are identical to those used for the same 
regions of the ultraviolet indexing card, and since they are given earlier 
in this chapter, they will not be repeated here. No code is involved for 
item (4) since the number is merely punched into these columns and the 
reserved sections of the card; items (9) and (10), have the same use as 
previously described. 

(1) Visible Absorptions. The coding of positions of visible absorption 
peaks is done in terms of wavelength in millimicrons. The coding resolution 
is 5 mp, that is, any number of peaks 5 m/u or farther apart may be coded 
individually. All peak values are rounded off to the nearest value ending 































222 


PUNCHED CARDS 


Column 

Punch 

Table 9-13 

Range (mp) 

Peaks 

11 

0 

350 to 450 

N one 

11 

1 

350 to 450 

One peak 

11 

2 

350 to 450 

Two peaks 

11 

3 

350 to 450 

Three or more 

11 

5 

450 to 550 

None 

11 

6 

450 to 550 

One peak 

11 

7 

450 to 550 

Two peaks 

11 

8 

450 to 550 

Three or more 

12 

0 

550 to 650 

None 

12 

1 

550 to 650 

One peak 

12 

2 

550 to 650 

Two peaks 

12 

3 

550 to 650 

Three or more 

12 

5 

650 to 750 

None 

12 

6 

650 to 750 

One peak 

12 

7 

650 to 750 

Two peaks 

12 

8 

650 to 750 

Three or more 

13 

0 

750 to 850 

None 

13 

1 

750 to 850 

One peak 

13 

2 

750 to 850 

Two peaks 

13 

3 

750 to 850 

Three or more 


in 5 or 0 before coding. Wavelength intervals covered by each column are 
printed at the head of the column. The “0” punch value is that printed at 
the head of the column and each successive digit in the column represents 
an increment of 5 m#i. Thus, a value of 560 m/i is coded by a punch at the 
2 position in column 5 which is headed by the number 550. Overpunch 
codes used in this section are the same as used on the ultraviolet card. 

(2) Number of Peaks. The number of peaks in each 100 m/i interval of 
the spectrum are indicated by punches in columns 11, 12 and 13 according 
to the direct code given in Table 9-13. Thus, a spectrogram that exhibited 
peaks at 375, 560, 575, 645 and 770 my. would require code punches at 
11-1,5, 12-3 and 13-1. 

(3) Intensity of Peaks. The intensity, in terms of absorbence for a solu¬ 
tion of 1 gram per liter in a 1-cm cell (the absorptivity), of the strongest 
peak in each 100-m/x interval is punched into columns 14 through 18. The 
intensity code used is the same as provided for coding the intensity of ultra¬ 
violet absorption peaks. (See Table 9-11) The range code for the visible 
card follows: 


Column 

Range (mji) 

14 

350 to 450 

15 

450 to 550 

16 

550 to 650 

17 

650 to 750 

18 

750 to 850 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 223 


Thus, a peak at 625 mu having an intensity of 25 would be coded as a 3 
punch in column 16. 

(11) Reference or File Number. As with the other A.S.T.M.-sponsored 
cards, the serial number of the spectrogram is punched into columns 73 
and 78 and the type of data coded into the card is indicated by punches 
in column 80. The source of the visible data are coded into column 79 as 
follows: 


79-A—(To be assigned) 

79-B—User’s own file of spectra 

79-C—Spectra issued by the NRC-NBS Committee 

79-D—Spectra abstracted by A.S.T.M.-sponsored groups. 

The column 80 code is given in the section on infrared. 

Mass Spectrometry 

The mass spectrum of a compound provides a unique set of data which 
can be used for qualitative analysis. In part, such an analytical operation 
involves a comparison of mass spectral data obtained from the unknown 
material with that obtained from known standard materials. Such a mass 
spectrum is rather complex and is usually represented by the actual trace 
from a recorder, a schematic drawing or a tabulation of the various mass- 
charge ratios and relative intensities. A listing of certain other operational 
factors essential to the production of comparable data usually accompanies 
such mass spectra. A large library of mass spectra and a good means of 
sorting and indexing it are essential to effective and efficient qualitative 
analysis. The ever-growing accumulation of mass spectral data available 
from the American Petroleum Institute Research Project 44 provides such 
a library and many laboratories have made notched card files of this and 
other data to facilitate the necessary matching operations. 

Hand-Sorted Cards. Two systems employing Keysort cards to facili¬ 
tate sorting of mass spectral data have attracted considerable attention. 
They are sponsored by two manufacturers of mass spectrographs, namely, 
Consolidated Electrodynamics Corporation and General Electric Com¬ 
pany.* ** Since the C.E.C. system has been incorporated into cards that are 
commercially available 2S , it will be described in some detail. 

The Consolidated Electrodynamics Corporation card (Figure 9-8) pro- 

* Thanks are due The Consolidated Electrodynamics Corporation, Pasadena, 
California, for permission to reproduce in this chapter its copyrighted mass spectrum 
card. 

** “Keysort File of Mass Spectra,” Consolidated Electrodynamics Corporation, 
Pasadena 8, California. 



224 


PUNCHED CARDS 



Figure 9-8. Consolidated Electrodynamics Corp. Mass Spectrum Card. 


vided for notching the following information: 

(1) Molecular Weight 

(2) Boiling Point 

(3) Elements 

(4) Ion Mass of Peaks 

All the data available concerning one compound are punched into or printed 
on a single card. This includes, in addition to the data listed above, the 
name and formula of the compound, source and purity, type of instrument, 
accelerating voltages, serial numbers and other pertinent data. 

(1) and (2) Molecular Weight and Boiling Point. These values are notched 
into the designated areas of the card by means of the familiar, 1, 2, 4, 7 
system. Provision for molecular weights to 999 is made and boiling points 
are punched in at 10°C intervals with only the tens and hundreds digits 
being notched. A special position is reserved to indicate that the number 
punched has a negative value. 

(3) Elements. The identity and a rough indication of the number of the 
common elements of organic chemistry are coded into one section of the 











QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 225 


card according to the following schedule of shallow and deep punches: 


Number of Atoms 


Element 

Shallow 

Deep 

Hydrogen (H) 

1-12 

13 or more 

Carbon (C) 

1-4 

5 or more 

Halogen (X) 

1-2 

3 or more 

Sulfur (S) 

1-2 

3 or more 

Nitrogen (N) 

1-2 

3 or more 

Oxygen (O) 

1-2 

3 or more 

Misc. (M) 

1-2 

3 or more 


(4) Ion Mass of Peaks. All the rest of the holes in the card are devoted to 
recording the ion mass of the largest or most distinctive peaks in the mass 
spectrum. The parent mass peak is always included if it is 8 per cent or more 
of the base peak. The shallow punch positions code each mass value from 12 
through 100. Deep punch positions carry the individual mass values up to 
150. Thereafter, there is one punch position for each two mass values from 
151-152 through 170-180, and for every ten masses from 181-190 through 
371-380. Special holes code peaks between 381 and 400, 400 and 600, and 
600 and 800. These values are all printed on the card to identify the proper 
holes. 

Conventional Keysort sorting operations are used to arrange cards in any 
one of several possible orders, or to search for particular cards having 
specific sets of data for matching and identification purposes. A complete 
appreciation of what can be done with these, or any of the other punched 
card systems, can be had only after actually using the cards for some time. 

The system developed at General Electric 26 , 27 is quite similar to the 
C.E.C. method just described. There is no provision for boiling points and 
although space is allotted for the coding of the elements in the compounds, 
it has not been used as yet. Punch positions are provided for indicating 
whether the data were obtained from a compound, a pyrolysis product or a 
mixture. A unique feature of the G. E. card is the affixing of the actual re¬ 
corder tracing of a rough spectrum to the card. 

Machine-Sorted (IBM) Cards. A system for indexing mass spectral 
and chemical structure data into IBM cards for sorting and correlating pur¬ 
poses was proposed by workers at Wyandotte Chemicals Corporation. 24 
This card was identical to the previously described Wyandotte-A.S.T.M. 
card for indexing infrared absorption data, except that provision was made 
for indexing the mass spectrum peaks, the strongest peak and the molecular 

** Zemany, P. D., “Punched Card Catalog of Mass Spectra Useful in Qualitative 
Analysis,” Analytical Chemistry, 22, 920-22 (July 1950). 

n Zemany, P. D., “Identification of Complex Organic Materials,” Analytical 
Chemistry, 24, 1709-13 (November, 1952). 



226 


PUNCHED CARDS 


weight. The further development of the card and system was assumed by 
Subcommittee IV of A.S.T.M. Committee E-14 on Mass Spectrometry. 
There resulted a detailed proposal by workers at M. W. Kellogg Company*® 
which increased the punching resolution, provided for both a base and 
parent peak, incorporated a more complete molecular formula and modified 
the chemical structure codes to meet the more limited classes of compounds 
susceptible to being handled in mass spectrographs. This was followed by 
a proposal by workers at Dow Chemical Company 29 which, although it 
makes use of IBM cards, is designed as a handsort or search file. The card 
carried the serial number, molecular weight, boiling point, number of 
chlorine and bromine atoms together with the mass numbers of the ten 
highest peaks, five other peculiar or particular peaks and as many as seven 
highest fractional peaks. Then, as many copies of each card are made as 
there are peaks punched into it. The cards are sorted and collated into 
blocks containing cards that have common mass numbers, then within each 
mass number block the cards are arranged according to the relative height 
of the peak on the particular card, and finally the blocks are arranged in 
order of the mass number. Thus, a copy of each card will be found in each 
mass number block for which it has a coded peak. With such a file one can 
go directly by hand to extract all cards indexing compounds whose highest 
peak has a given value or can also include all compounds that have an 
indexed peak at the given value regardless of its relative height. This 
achieves the results usually obtained by a first sorting operation at the ex¬ 
pense of increasing the number of cards in the file by a factor of 10 or more. 
However, the comparison of cards in a given block to correlate the several 
peaks in a given spectrum with a given compound, when as many as 15 
peaks may be involved, could become a rather complicated hand-sorting 
operation. 

Since A.S.T.M. Committee E-14 has not taken official action on any sys¬ 
tem involving IBM cards and none are commercially available from other 
sources, a detailed description of the codes and procedures of such systems 
as are under consideration is not advisable for time may soon render any 
one obsolete. 

Empirical Formula -Name Index 

With the rapid accumulation of thousands of Wyandotte-A.S.T.M. cards 
indexing the absorption spectral data and chemical structure of as many 
different chemical compounds, the problem of maintaining an alphabetical 
index of the names of these compounds became rather complex. The need 

** McCrea, J. M., “A Proposed Indexing System for Mass Spectra,” submitted to 
A.S.T.M. Committee E-14, Subcommittee IV on May 26, 1953. 

*• McLafferty, F. W., and Gohlke, R. S., “A New Punched-Card Filing System for 
Mass Spectra,” presented at the A.S.T.M. Committee E-14 meeting on Mass Spec¬ 
trometry, New Orleans, La., May 28, 1954. 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 227 


for such an index resulted from the frequent desire to locate the spectro¬ 
gram of a given compound without having to resort to sorting the chemical 
structure data punched in the spectral data index cards. The complexity 
and duplicity in naming organic compounds made it desirable to establish 
a system that did not rely primarily on the name. Through the cooperative 
efforts of workers at Wyandotte Chemicals Corporation and Eastman 
Kodak Company, a system was developed which makes use of the empirical 
formula and a name punched into the same IBM card. Although IBM cards 
are used it is only for convenience in the initial preparation and subsequent 
duplications, since the cards are used as a hand file. The system, described 
in detail below, has been adopted by the American Society for Testing 
Materials and is being distributed by them, together with the other cards 
previously described. Thus, every spectral data card for a given compound 
has a formula-name card bearing the serial number of the spectrogram 
which serves to locate the spectrogram in the user’s files. 

Formula Name Cards. The formula-name cards indexing chemical 
compounds are designed to provide a ready means of obtaining all of the 
information about a given compound that has been coded into any df the 
other Wyandotte-A.S.T.M. cards. They may be arranged by machine into 
numerical order of the spectrum serial number, the numerical order of the 
empirical formulas, or into alphabetical order of the names and then used 
as a hand file for entry by any of these arrangements. The name, empirical 
formula and serial number are printed on the card to facilitate hand use. 
There is one card for each compound each time information concerning the 
compound is indexed into a different spectral data card. Thus, by entering 
the file for the compound benzene, one will find cards giving the serial 
number of the infrared absorption spectrograms, the ultraviolet absorption 
spectrogram and any other sets of data as have been incorporated into the 
system. 

The formula-name card (see Figure 9-9) is divided into the following areas 


/UUUUUUUUU'IJ 


HM fiji ywvwy 

• » ohi w « « it a 

1 1 ill 1 1 1 1 ill 1 ill 


lllllllllllllllttltlllltltl 


nnURnrij 

lira a 


llllllll 

i I m II 11 
11111111 

1222222 ill 222222222222 2| 
nmiisnmnnnml 

^UNCMCO Ca *0 

4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4| 

0<«TMiauTCO 

111»s»i j j s i m i s m s s s| 

umkm Mem im rittiM wtitmi 

I 11 111111 I I 11 I I « I I I I I l| 

>••• l*C( «TMUT 
111111111111111)111111] 
VMIkAOCwriM a. »|NNtVkV*NI* 

1111111111111111111111 

tiniiiniiiiiiiiiiiii 




ill • i • i • i i 


ISISISI) 

44444444 

SSSSS9SS 

llllllll 

1111 9 11 


III 


I I I 
222 

22) 

444 

III 

III 

777 

III 

III 


11111111111111111(111111111 

2222222227222222222222222221 


11111 


12 2 2 2 212 7(2 212 2 2 2D 


7)2)2 3 minus 2)2) JJ 221)1 nmilllUSM 

▼Mia moik 

4 4 4 4 4 4 4 4 4 4 4 I 4 4 4 4 4 4 4 4 4 I 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 

w*a **c»*»io -m th* 

unnusususnuuunuistusiiimi 

»H»eci ««ac**CM 

•lllllllllllllllllllllllllllllllllllllll 

MU*KCH AMO OI»lkO»MCNT Orwia>OM 
1111111)11111111111111111111111111111111 
WVAMOOTTK CMfMC«k( c OMO* ATlOM 

iii iiiiiiii ii i iii i ii ilium mu min 

WVANOOTTI. MCMilAM 

mill iiiiiiii iiiiiiiiiiiiiiiiiiii t i iii 




ATi«a. atMMMI. (KMIl 


Figure 9-9. Wyandotte-A.S.T.M. Name-Formula Card. 































228 


PUNCHED CARDS 


for incorporating data: 

(1) Elements—columns 1 through 8 

(2) Empirical Formula—columns 9 through 22 

(3) Miscellaneous code—column 25 

(4) Compound Name—columns 26 through 65 

(5) Reserved by A.S.T.M.—columns 23 and 24 

(6) Reserved for Private Use—columns 66 through 70 

(7) Reference of Serial Number—columns 71 through 80. 

Each of the areas will be discussed in sufficient detail to permit one to 
make general use of the cards. Additional information may be obtained 
from the A.S.T.M.* 0 

(1) Elements. The identity of every element in the compound being 
coded is indicated by punches in this section. The same elements code used 
on the Wyandotte-A.S.T.M. x-ray diffraction cards and on the inorganic 
section of the chemical classification code for the infrared absorption data 
cards (see Table 9-5) is employed here in column 1 through 8. Thus, the 
numbers in brackets give the column numbers, and actinium would be 
punched at the “0” position in column 1, etc. This section can be used to 
segregate cards according to particular elements. 

(2) Empirical Formula. Columns 9 through 22 are used to record the em¬ 
pirical formula of the compound being indexed. These numbers are punched 
directly into the appropriate columns and then interpreted or printed by 
machine along the top of the card. The chemical symbol for each element is 
also printed on the card for ease in hand searching. Table 9-14 relates the 
elements involved, the columns and the printing positions for interpreting 
the numbers in the proper place on the card. 

It will be noted that only the more common elements are included in the 
table. Numbers of atoms greater than can be punched into the columns 
provided are recorded as the highest number that can be punched. Poly¬ 
mers and indeterminate structures receive no empirical formula punch. The 
empirical formulas punched here include all elements involved in the com¬ 
pound except water of hydration. Salts such as aniline hydrochloride, 
sometimes recorded as C«HjNHj-HC 1 would be punched into the card 
as CgHgClN. An additional code in the next section of the card has been 
provided to indicate that such salts are involved. 

(3) Miscellaneous Code. Column 25 provides for coding salts where the 
information would not be apparent from the combined empirical formula 
given on the card, it indicates when the compound coded is inorganic as 
well as the presences of water of hydration, which is not included in the 

*° “Codes and Instructions for A.S.T.M. Empirical Formula-Name Index Cards,” 
Ibid, (1956). 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 229 


Table 9-14 


Column 

Element 

Printing Position 

9-10 

c 

1-2 

11-12 

H 

4-5 

13 

Br 

7 

14 

Cl 

9 

15 

F 

11 

16 

I 

13 

17-18 

N 

15-16 

19-20 

0 

18-19 

21 

s 

21 

22 

Si 

23 


empirical formula. The miscellaneous code follows: 

25-y HC1 
25-x HBr 
25-0 HjSO« 

25-1 Acetate 

25-2 Oxalate 

25-3 Phosphate 

25-4 Ammonium 

25-5 Nitrate 

25-6 H t O (hydration) 

25-7 Inorganic 
25-8 

25-9 Other Acid Salt. 

(4) Compound. Name. The name of the compound, as closely as can be 
approached within the limitations of standard IBM equipment, is punched 
into columns 26 through 64. Since only capital letters, digits, comma, dash 
and slanting line are normally available, the names are printed in a modi¬ 
fied but readily recognizable form in most cases. The “inversion” naming 
system* 1 as used by Chemical Abstracts is favored for use on these cards. 
However, no attempt has been made to rename all compounds by the 
Chemical Abstracts System. Such names as were supplied by authors were 
merely rearranged, applying the Chemical Abstracts principles, so that an 
“index name” could be used for alphabetizing. With these cards the 
empirical formula is all important and the name can be considered trivial. 
One need only be able to recognize any one of the possible names of the 
compound being searched for. 

The first letter of the “index name” is always punched into column 30. 
Columns 26 through 29 provide for digits and/or letters which usually 
precede an index name but take no part in determining the alphabetical 

11 "The Naming and Indexing of Chemical Compounds,” Chemical Abstracts, 39, 
No. 24 (Introduction to the 1954 Subject Index). 



230 


PUNCHED CARDS 


sequence. If there are more such characters than can be accommodated 
in four columns, they are placed at the end of the name followed by a 
dash. Thus, 1,4,5,8-Naphthalenetetrol, 3-chloro is printed as NAPHTHA- 
LENETETROL, 3-CHLORO-l ,4,5,8-. Greek letters are either spelled 
out or are represented by English equivalents for economy of space. The 
“prime mark” (') is indicated by the letters “PR”. Thus, o,o'-Biphenol is 
printed as BIPHENOL, 0,0PR-. Such other abbreviations as “C” for cis, 
“T” for trans, “M” for meta, “D” for dextrorotary, “N” for normal and 
many others which are perfectly obvious, are used in printing the names. 
Since parentheses are not available on the regular Model 552 interpreter, 
a slanting line has been used. Thus, a name written as 2-(methyldithio)- 
ethanol is printed on the card as ETHANOL, 2-/METHYLDITHIO/-. 
The same slanting lines must serve also as brackets. Every attempt is made 
to make the names as readable and correct as possible. The name is in¬ 
terpreted or printed in the lower printing space on the card in positions 5 
through 44. On these cards an “x” overpunch produces the comma (,), a 
“y” overpunch produces the dash (-) and a combination of 0 and 1 punches 
in any column produces the slanting line (/). 

If the name is too long to be punched into the 39 columns available on the 
first or parent card, it is broken at a normal position and the rest punched 
into the same columns of a second, or trailer, card which carries the same 
serial number. When this is done, a “T” is punched into column 65 of the 
parent card and a 9 into column 65 of the trailer. If the name cannot be 
punched into two cards, then a second trailer (or a third) may be used in 
which case both the letter T and the digit 9 are punched into column 65 of 
the middle trailers and only 9 into column 65 of the last trailer. All trailers 
carry the same serial number of the parent card but are different in color 
and carry no other punches or printing other than the portion of the name 
and the serial number. Punches into column 65 are interpreted to upper 
printing position 38. 

In normal use, the cards are arranged strictly in numerical order of the 
number of atoms, and working from left (carbon atoms) to the right as 
printed across the top of the card. When no atoms of a particular kind are 
present, the fact is ignored and the next element to the right determines the 
order, but such cards are all placed behind the cards with formulas that do 
contain the element. Thus, the file begins with compounds containing one 
C and one H atom and all compounds containing one C and no H atoms fall 
behind those containing one C and the highest number of H atoms. This 
system is adhered to strictly. Otherwise, when atoms are present, the num¬ 
ber of such atoms determines the position of the card and, working from the 
left, all cards having a given number of atoms are added to the file before 
cards containing a higher number of such atoms are included. All com- 



QUALITATIVE CHEMICAL ANALYSIS BY SPECTRAL METHODS 231 


pounds that do not contain carbon fall behind all compounds that do 
contain carbon and since this former group is chiefly inorganics it has been 
arranged in alphabetical order of the names. Polymers, trade name ma¬ 
terials and all compounds that have no empirical formulas are filed alpha- 
tically by name in the last section behind the inorganics. 

A brief examination of the cards as they are distributed serves to familia¬ 
rize one with the system. In this arrangement of the cards, it is convenient 
to file the trailer cards separately from the parent cards since they bear no 
empirical formula data, and to keep them in numerical order of the serial 
number so that they may readily be located when necessary. 

(5) and (6). Reserved Areas. As on all other Wyandotte-A.S.T.M. cards, 
certain columns are reserved by A.S.T.M. for future use and another sec¬ 
tion is set aside for private use by individual laboratories. In the Formula- 
Name cards columns 23 and 24 are reserved by A.S.T.M. for their own 
purposes and columns 66 through 70 are available for private use. 

(7) Reference or Serial Number. The serial numbers punched into these 
columns are the same as those in the corresponding infrared, ultraviolet or 
visible data cards for the compound. This includes the letters in columns 
79 and 80 so that the designation on the card provides a direct reference 
to the location of the spectral data in the literature or local files. 

Acknowledgment 

The author wishes to thank Wyandotte Chemicals Corporation and the American 
Society for Testing Materials for permission to publish major sections of this work. 



Chapter 10 

AN APPLICATION OF RANDOM CODES 
FOR LITERATURE SEARCHING 


Claire K. Schultz* 

Librarian, Merck Sharp & Dohme Research Laboratories 
West Point, Pennsylvania 

Introduction 

The random coding technique for indexing journal references has been 
employed in the Sharp & Dohme library since 1950. The library now in¬ 
dexes about 15,000 articles per year; a small amount with respect to the 
needs of some of this book’s audience, but probably an “average” volume 
for special libraries attempting to index current literature for their organ¬ 
izations. 

This library’s literature service has to satisfy a group whose interests 
touch on nearly every phase of the biological, medical, and chemical 
sciences. The technique evolved for coding the names of diseases affecting 
man and domestic animals is quite specific, as is that for coding organic 
chemicals of known structure. Additional subject description for a given 
paper is supplied from a nonclassified, alphabetically arranged list of sub¬ 
ject words. 

The system can still be regarded as an experiment, in that ways for im¬ 
proving it are always under consideration. The virtues of a system that 
can be changed without invalidating any previous input must be recognized. 

A staff of three typists and two professional people handle the input and 
output of the system. 

The Conception of the System 

The first consideration of punched cards for library use made it clear 
that a change from conventional indexing to a punched card system would 
be desirable only if the new system could provide all the functions of a 
standard index and also offer significant advantages. A system was needed 
that was capable of storing enough information to define and describe a 
reference (a standard index can do this) and that could also retrieve infor¬ 
mation quickly and easily by associating any and all subject fragments 
presented at the time of a search (a standard index cannot do this in many 
instances). 

* Present address: Univac Division, Sperry Rand Corp., Philadelphia, Pa. 


232 



RANDOM CODES FOR LITERATURE CODING 


233 


The punched card system adopted by the library had to be capable of 
handling a considerable volume of references, in terms of standard index 
systems, and also of meeting the diversified subject needs of the scientists 
it served. It was felt that in designing the library’s application of punched 
cards an attempt should be made to get as much information as possible 
on one card, in order to make correlation of that information as easy and 
as meaningful as possible. It was recognized that the use of random super¬ 
imposed symbols needed less card space than any other coding technique. 

Continuing this reasoning, the point was reached where experiments 
could begin. Journal articles were coded from a dictionary of subject words 
arranged in alphabetical order. Each word had been assigned a random code 
number that represented four holes in a punched card, the codes for all of 
the subject words were superimposed in a field of ten columns on the card. 
An additional ten columns were used for coding the journal name, author 
and date of the reference. At that time IBM could not offer a machine to 
search for random superimposed codes, so a Remington Rand Sorter had 
to be used. 

The enlarged and refined punched card system in operation at present, 
employing an IBM 101 Statistical Machine for searching (Figure 10-1), has 
grown from the base set by this 1950 experiment. 

Random Codes 

Mathematical discourses on coding systems, including random codes, 
are to be found in the literature 1 • *. A few lay observations stemming from 
the application of random codes will be presented here. 

It has been pointed out that the technique of superimposition of random 
symbols offers the advantage that many “bits” of information can be 
coded into only a few columns of a card. Inherent in this technique is the 
possibility of creating false selections. There are numerous factors that 
modify this latter fact, some of which can be mentioned here. 

The number of punches assigned to a code is one of the basic considera¬ 
tions. The more definitive the code, i.e., the more punches assigned to a 
code, the less probability there is of synthesizing it by chance when search¬ 
ing. However, the more punches used to define a term, the fewer the num¬ 
ber of terms that can be superimposed into a field before it becomes satu¬ 
rated 1 . This often has definite practical significance. Also the longer the 
code, the more cumbersome it is to work with in the clerical sense, and thus 
the greater the amount of human error that can be expected to enter the 
system. 

1 Calvin N. Mooers. “Putting probability to work in coding punched cards—Zato¬ 
coding.” Presented before the Division of Chemical Education at the 112th meeting 
of the American Chemical Society, New York, Sept. 15-19, 1947. 

* Carl 8. Wise. Mathematical analysis of coding systems (Chapter 21, This Book). 



234 


PUNCHED CARDS 



Figure 10-1. Electronic statistical machine Type 101, with auxiliary dial board. 

Taking all of this into account, a code of four punches was decided upon 
for the application being described. Four-punch codes allow the superim¬ 
position of up to sixteen terms into a field of 100 punches. According to 
theory 1 , not more than 69 per cent of the holes in any random field may 
be used up in coding information into that field. In this application a field 
contains 100 punching positions,- so 69/4 or 17 and a fraction terms, then, 
could be used per field. Actually, this number of terms becomes a little 
larger due to the overlapping of codes; for example, if the code 01-17-(49)-92 
is punched and another term has the code 05-27-(49)-81, only threenoles 
instead of four are needed to punch the second code into the card. How¬ 
ever, to operate with a margin of safety, an upper limit of 16 terms per 
field has been observed here. In practice, the resulting false selection is 
small enough to cause insignificant interference. It is a natural practice to 
combine a group of terms for searching, rather than to look for a single one. 
This practice of amalgamating codes into a search “pattern” is an impor¬ 
tant contribution to the elimination of false selections. 

Not all codes, though, are equally selective. The fact that they overlap, 
some by one digit, some by two digits, and some by three digits, makes them 
differ in their selectivity. Then, too, there is the fact that not all of the 
codes chosen for assignment to a dictionary will be used equally often. Some 
terms or groups of terms are needed more frequently than others for both 
indexing and searching. The codes for terms that are often used cut down 


RANDOM CODES FOR LITERATURE CODING 


235 


the selectivity of codes having numbers in common with them. Illustration: 
In this library the terms: 

humans 

therapy 

experimental 

animals 

occur frequently as a group on the punched card. The digits in their codes 


08-10-36® 
97-86® 78 
14®67-76 
Jl-52-55 



synthesize many other codes, one of which is traced in the above illustra¬ 
tion. Searching for a term with this code would be almost useless because 
the volume of cards bearing these four codes would drop as false selections 
in prohibitive proportions. To alleviate that situation in this library, the 
terms used most frequently, either singly or in combination, were removed 
from the subject field and given direct punches in an otherwise unused por¬ 
tion of the card. Figure 10-2 shows these words and their punching positions. 
They can still be selected in combination with other codes used in the 
system but they no longer cut down the selectivity of other codes. 

For the application being described, the list of random numbers assigned 
to the subject dictionary was derived from a table of random numbers in 



Figure 10-2. Literature reference card before reference is entered. 




236 


PUNCHED CARDS 


r CX3C3CXXXXDCX3CXXX 
•—!-1-* !_»_i_!_I_I 8 !! 1—M_ 


XXXXXXXXXXXX5 

ti n »» it «t w w n n w f» ft »> 


c*rx:lDC»>:>ft3C«5oi5cia5cl|5aD5c»iA5cL ^iAdcl J:0^c0 


| CftDCUD<^>Iil-C4!3C5l3C|JDC7l^C|1DCf|Z ^DCKU^e^CMplDCl 

g c«x:ir3C»Dc3l^c«>:»3cC>:Tr>^>nr4^^ N ^ c ^ N 4 r ^^4 r 2 :x:: 2^:?^c24 r ? DC 2>^2^?^?^c:?^c?^ 


| c|pnjDC?j>rir>c<pCM^^cn>:iJ>n44:0^c«4-t)Dc«4 3^3 

c«t>ri( Dot jcK^c«cxl4>^C3Cl(DCtrx^|^ 3CP=^C dcp ^i4 dc4 
e*pctv 



3Dc3Dc3Dc3ifc3Dc3Dc3Dc3Dc3^c3Dc33 
4 x<dc4dc<^4x<x4x4x4x4x43 
5dc5dc5dc5ix=5dc5dcS3 
3c|x$x|x$x6x43 

CRDCllDCllDCjTDClTDCjrX«3C7T3CaiDC*=^MDCT=(^DCT^:73C7^:70C7DC72C7=j=73C;DC73C7DC73C70C;D 

X|X|XSX|X|X|3 


^fDcn jzix^DcixSxy 
**« 


OdcOdcOdcO 

JOURNAL 

:\x\x\x\ 


:0dc0^'0 0 '*03c0>=0 

MltAl M. 

: 13Cl Dcr^cl3cf3cl^c 13 


•uu. 




Figure 10-3. Face of mark sensing card. 


1 

\ 


\ 


CXXXXXXXXXXXX X X X X X X X 

A 

CXXXXXXXXXXXXXXXXXXXXXX 


X X X X X X 


txtxtdtt. 


uxnxmxmxuxm>--i\^m> *4* 

%txaxsx«x«xcxii ■ 

»x ux 



c7oc73c7=fc; 


izt 


DC 

DC 


:2DCjDC3DCjDC2^CtDCr 
|3 C|dc3 xtJdcJdcIdc* 

4x4x4x|x4x4xi 

xSxJxSxJxJ: 
44x(x|x|x(x04 


*®t t 


:1x}x}xJx1xJxJ. 

i|x|x|x|x|x|xti 


|:|3ClDClx;lD 

;x}x}xj3 

JXJX1XJD 

4x1x4x43 

SdcSdcSdcSd 

DC(DC$DC$D 


Jx_;dc7dc7d 

|X|X|X|3 

±l2£l2£l2£l2. 


qPf^q w rij*, ~ ilxtxfxtxlaixltlxlxtxt 


Figure 10-4. Back of mark sensing card. 


Fisher and Yates 3 . That table lists the last 10 digits of a 20 place loga¬ 
rithmic table. Establishing a code for a field of ten columns (100 punching 
positions) allows the use of the numbers 00 through 99. (See the first 10 
columns of Figure 10-3.) Constructing the code designating four punches 
in the field, therefore, required 8 digits. The first 8 digits of each entry in 
the Fisher and Yates table were utilized so that the entry 1324354657 
yielded the code 13-24-35-46. The listings in the table were assigned to the 
alphabetically arranged list of subject headings, with no regard for estab¬ 
lishing numerical relationships among the subject words. 

Searching Random Codes by Machine 

In the course of this library’s program, experience has been accumulated 
with both the Remington Rand Sorter and the IBM Statistical Machine, 

J R. A. Fisher and P. Yates. Statistical tables for biological, agricultural and 
medical research. London, Oliver and Boyd, 1938. 








































RANDOM CODES FOR LITERATURE CODING 


237 


Type 101. The operational aspects of these machines that are important to 
a system applying random codes can be summarized briefly. 

The Remington Rand equipment comes with a sorting block that covers 
144 punching positions (12 columns of the card). Various types of sorting 
pins make it possible to select a pattern of numbers within that area or to 
reject a pattern that might be associated with the pattern being selected. 
Selection into pockets is controlled by a bridge operating over only one 
column in any one pass through the machine. The decision as to what 
pocket will bear the product of the search, then, has to be a function of the 
pattern being searched. The Remington Rand card has 540 punching posi¬ 
tions. The searching rate is 25,200 cards per hour. The machine will search 
for any pattern of punches that might be put into the 144 positions covered 
by the block. Aside from the electric motor driving the machine, it is com¬ 
pletely mechanical in its sensing and selecting operations. 

The IBM machine will search for a pattern of up to 60 holes anywhere 
on the card. Sequencing (either alphabetical or numerical) by means of 
the IBM 101 is achieved more easily and quickly than with the Remington 
Rand equipment; the machine also counts and prints. The sorting speed of 
the machine is 27,000 cards per hour. However, preference for the 101 has 
been based primarily on the increased amount of correlation possible at 
the time of the search. 

To demonstrate this, one needs to consider the use of logical connectives 
in punched card sorting. A variety of logical patterns may describe the 
relationships of terms being used to formulate a search. For purposes of 
demonstration, the terms being searched can be designated as A, B, C, D, 
and can be assigned meaning as exemplified in the following: 

A. Benemid (therapeutic agent) 

B. Penicillin 

C. Gout 

D. Pneumonia 

The logical connective and between each would signify that only those 
papers with reference to Benemid and penicillin being used in therapy of 
both gout and pneumonia are wanted (A + B + C + D). This type of 
searching, i.e., the use of the logical connective and can easily be accom¬ 
plished by any punched card system. 

If the request is for references to the therapy of pneumonia by these 
drugs, but not if the reference also concerns gout, the relationship is: 

A + B + D - C 

The use of the logical connective but not is accomplished in the Remington 



238 


PUNCHED CARDS 



Figure 10-5. Close-up of an auxiliary dial board. 


Itand sorter by using reject pins in the sorting block. With a hand-system, 
this type of search can be done only by selecting all four, 

A + B + C + D 

and then removing C from the pack selected in the first sort. The simple 
way in which the IBM accomplishes the search A + B + D — C will be¬ 
come clear as the discussion progresses. Other logical connectives that can 
be dealt with only by an electronic sorter are exemplified by: either-or ; 
if-also; and-if. 

The IBM equipment is so flexible that all of the combinations represent¬ 
ing the logical connectives among the terms being searched are readily 
separable in routine operations. To facilitate these separations a wiring 
system has been developed to deliver the 15 possible combinations of 
A B C D each time a search is made 4 . The use of a dial board obviates the 
time and technical training needed to wire a control panel for a search, 

* The principle which led to this development of the auxiliary panel board was 
conceived by Mr. Bruse Moncrieff and his associates at the Home Office of the Pru¬ 
dential Life Insurance Company, Newark 1, New Jersey. 












RANDOM CODES FOR LITERATURE CODING 


239 


and through its use, it makes the 101 a more practical and efficient tool for 
literature searching. The dial board allows the codes for A, B, C, and D to 
be set in without requiring special knowledge or skill. Figure 10-5. 

To continue with the example given above, the method of wiring used 
in this library delivers the answers to more questions than the specific one 
being asked; the answers to corollary questions are ready-made. An exam¬ 
ination of what has dropped into each of the pockets of the machine as a 
result of this search will point up this fact. 


Pocket 

Combination 


12 

ABCD 

Papers making reference to Benemid, Penicillin, Gout and 
Pneumonia. 

11 

BCD 

References to Penicillin, Gout and Pneumonia but not Benemid. 

10 

ACD 

References to Benemid, Gout, Pneumonia, but not Penicillin. 

1 

ABC 

References to Benemid, Penicillin, Gout, but not Pneumonia. 

2 

ABD 

References to Benemid, Penicillin, Pneumonia, but not Gout. 

3 

CD 

References to Gout and Pneumonia, but not to Penicillin or 
Benemid. 

4 

BI) 

References to Penicillin and Pneumonia, but not to Gout or 
Benemid. 

5 

BC 

References to Penicillin and Gout, but not to Benemid or Pneu¬ 
monia. 

6 

AC 

References to Benemid and Gout, but not Penicillin or Pneu¬ 
monia. 

7 

AD 

References to Benemid and Pneumonia, but not Penicillin or 
Gout. 

8 

AB 

References to Benemid and Penicillin, but not Gout or Pneu¬ 
monia. 


A 

References to Benemid when it was not used in combination with 
Penicillin and when neither Gout nor Pneumonia were men¬ 
tioned. 

9 < 

B 

References to Penicillin when neither A, C, nor D were present 
in the paper. 


C 

References to Gout when neither A, B, nor D were present. 


D 

References to Pneumonia when neither A, B, nor C were present. 


The last four can be separated by passing the cards from pocket 9 through 
the machine again. 


Dictionary 

Important as they are, the three elements of a punched card system— 
the equipment for sorting, the design of the card, and the type of number¬ 
ing system employed—can be looked on as tools for putting a subject 
dictionary into effect. Without a well constructed dictionary, the full val¬ 
ues of punched cards for indexing literature cannot be realized. 

There are two basic approaches to a dictionary with which to begin one’s 
thinking: (1) an ordered classification in which the terms used and the 



240 


PUNCHED CARDS 


numbers assigned to them are correlated and dependent or (2) a nonclassi- 
fied system where the dictionary is developed without regard to the logical 
relationships of terms. Example: 


Classified Dictionary 

animals 

mammals 

dogs 

rats 

virus diseases 
mumps 


Nonclassified Dictionary 

animals 

dogs 

mammals 

mumps 

rats 

virus diseases 


For the classified dictionary, code numbers are usually assigned to empha¬ 
size logical relationships among entries, e.g.: 


animals 

2 

mammals 

2.2 

dogs 

2.21 

rats 

2.26 

virus diseases 

5.0 

mumps 

5.4 


In a nonclassified random dictionary the entries are mutually independent 
and the assigned codes are of equal weight: 


animals 

dogs 

mammals 

mumps 

rats 

virus diseases 


14-35-27-48 

16-19-88-92 

01-08-36-99 

02-08-16-31 

33-47-66-84 

52-56-68-72 


Terms may be incorporated into a non-classified dictionary as needed be¬ 
cause there is no difficulty in making additions. There may be some diffi¬ 
culty in adding to a classified dictionary because its scope is more or less 
defined at the time it is set up. If one decided the next year to do research 
in three or four large fields not within the original scope of the system, the 
classified dictionary is likely to be under great stress. 

One of the first things that becomes obvious in starting a non-classified 
dictionary is that it must never repeat any word or phrase. Classifications 
such as Dewey®, or Library of Congress 9 , use: 

Religion—history 
Medicine—history 

‘Melvil Dewey. Decimal classification and relative index, 14th ed., New York, 
Lake Placid Club, 1942. 

0 Martin, Nella Jane, ed., Subject Headings used in the dictionary catalog of the 
Library of Congress, 5th ed., Washington, Library of Congress, 1948. 



RANDOM CODES FOR LITERATURE CODING 


241 


Quarterly Cumulative Index Medicus List of Subject Headings 7 uses: 

Penicillin-toxicity 

Sulfonamides-toxicity 

The repetition of the word history or toxicity has no use in a nonclassified 
dictionary. Any word appearing in a nonclassified dictionary can be used 
in any combination desired; it appears there only once. To search for all 
the information in the file on toxicity would be as easy as searching for all 
the information in the file on ■penicillin. This is certainly not true of a stand¬ 
ard card catalog where such a search would have to be carried through every 
drawer, toxicity being only a subdivision of the main headings. 

Since repetition is unnecessary, the over-all size of the dictionary is con¬ 
siderably reduced and every term appearing in it has a utility not to be 
found in any other type of authority list. A word such as antagonism or 
anti can be used to form: 

anti histamine 

anti spasmodic 

anti bacteria (antibacterial) 

anti sepsis (antiseptic) 

anti coagulation (anticoagulant) 


In developing such a dictionary, rigorous attention must be given to the 
exclusion of synonyms and closely related terms. After completing a search, 
one does not want to discover that he should have asked the machine to 
select the cards bearing codes for kittens and felines as well as cats. These 
entries must be cross referenced and must never be assigned separate code 
numbers. 

The present working dictionary in the system being described is a non¬ 
classified list of approximately 1000 indexing terms, consolidated as shown 
in the excerpt below: 


Random number Subject word 

12-15-29-91 ANTIHISTAMINES 
06-24-25-28 ANTIMONY AND 
ANTIMONY 
COMPOUNDS 
ANTIPYRETICS 
ANTISEPSIS 

36-46-54-82 ANTISEPTICS 

ANTISERUM 

ANTITOXIN 


Remarks 


use: Fever; Therapy 
use: Antiseptics 

also coded: Names of specific agents ap 
pearing in dictionary: Sterilization 
use: Anti; Serum; Immunity 
use: Toxin; Anti; Immunity 


7 American Medical Association. Quarterly Cumulative Index Medicus Subject 
Headings and Cross References, 2nd ed., Chicago, 1940. 


242 


PUNCHED CARDS 


This dictionary has thus far proved adequate for indexing some 30,000 
references. As seen in the example given, not every subject word listed has 
a code number. In many cases the subject word is expressed by a group 
of other coded terms as shown opposite ANTISERUM. The term itself 
does not have a specific code but is found by searching for the cards con¬ 
taining the codes for anti, and serum, and immunity. At first glance, this 
may seem cumbersome, but the dividends are to be found in considering 
searches for terms other than antiserum when the reference is desired. 
That is, if all of the references pertaining to immunity were desired, the 
cards referring to antiserums would be among those selected. 

In some cases terms used by themselves have less selective power than is 
true of individual terms in a standard index. With the present system, a 
word such as CELLS may be used with any body organ or tissue. BLOOD 
CELLS, PANCREATIC CELLS, or SERTOLI CELLS all have the same 
code number for CELLS, but are distinguished by the additional codes for 
BLOOD, PANCREAS, and TESTIS. The reason for this, again, is to make 
it possible to find the reference under a greater diversity of searching condi¬ 
tions. The indexer has to think generically and specifically about every 
reference handled if this system is to approach the ideal in usefulness. For 
a paper about the stomach, the indexer would use not only stomach (spe¬ 
cific), but also gastrointestinal tract (generic); the reason being that a re¬ 
searcher might be studying the effect of a certain drug on the gastrointesti¬ 
nal tract in general. With this type of coding he can get the papers on the 
subject without searching for esophagus, stomach, intestines, etc. It would 
be a mistake, though, to force the reader looking for references on the 
stomach to hand sort the entire pack of cards on gastrointestinal tract; 
references are conveniently indexed both ways, since to do so does not 
involve the preparation of more than one card. 

All of the thinking about the generic and specific relationships of each 
term has to be set forth in the dictionary if the indexing is to be consistent. 
If the indexer uses just penicillin one time in indexing an article and then 
antibiotics and penicillin the next, the file cannot be expected to yield all 
the papers on antibiotics when searched. Anyone used to thinking in terms 
of the standard indexing practice may forget to index antibiotics on the 
penicillin paper; the dictionary will remind him to do so when he looks up 
the code for penicillin. 

This same point could be made from many facets of the coding and 
searching. The indexer might analyze a paper very carefully and select the 
subjects indicating that the article gives information about the treatment 
of a disease in a child and that a certain dosage of a compound is given; but 
if the searcher wants all the papers on the use of that compound in human 
beings, he will not find the paper if it is coded only under children. It must 



RANDOM CODES FOR LITERATURE CODING 


243 


be coded under both humans and children and the dictionary must tell the 
coder: 


Children Also coded: humans. 

Humans Also coded: children, when pertinent. 

Thinking in terms of requests from the information file, searching will 
be simplified if the dictionary anticipates similarities of terminology as 
much as possible. One person might ask for a search on diagnosis of dia¬ 
betes, another for tests for blood sugar. These requests overlap somewhat, 
and if both words, diagnosis and tests, are coded by a different number, 
only a fraction of the wanted references will be obtained when searching 
for one of them. To take care of such words that are not synonyms, but 
which should not be given individual codes, our dictionary lists each of them 
in its alphabetical place, using the same code number for each of them. 

In the development of the dictionary, the Quarterly Cumulative Index 
Medicus List of Subject Headings provided a frame of reference. Q.C.I.M. 
had been used as the library’s authority list for two years previous to com¬ 
piling the punched card dictionary. All the terms in Q.C.I.M. were con¬ 
sidered for use with the punched cards, but the fact that the subject head¬ 
ings used by the library during the preceding two years had been checked 
made it easier to predict future needs. In addition to Q.C.I.M. coverage, 
many of the key research people reviewed an early version of the dictionary 
and suggested additional terminology which they felt necessary to cover 
their special fields of interest. The librarian correlated all of the suggestions 
and made the additions, cross references, and appropriate notes for the 
dictionaiy. 

This random number dictionary has always worked well within the limits 
set for it. It was found by experience, however, that the volume of drug 
names and disease states encountered in the references indexed by this 
library was larger than had been anticipated when the dictionary was 
built. The system was suffering from not being specific enough in these two 
areas. 

To correct the problem created by the drugs, it was decided to code them 
in a separate field from the other terms in the dictionary, and to assign every 
organic chemical of known structure a random number as it was encoun¬ 
tered in the literature*. This was begun in 1953. In the three years that 
have ensued, the number of individual compounds coded is a little less than 
one thousand. The methodology for using the “chemical field” is the same 
as for the “subject field.” In fact, the same group of random numbers was 

* This statement needs to be modified to some extent. When it seems more expedi¬ 
ent, salts are coded as the parent acid. In this sense, a random number can denote a 
group of compounds. 



244 


PUNCHED CARDS 


reused for assignment to compounds. Because they occur in separate fields, 
this is workable. 

Diseases could have been taken out of the original dictionary and handled 
in the same way as compounds, or they could have been elaborated more 
thoroughly within the original dictionary. An intracompany development, 
though, made a third choice more expeditious. It was decided, for the sake 
of better intracompany communication, to adopt the American Medical 
Association’s Standard Nomenclature of Diseases 8 , used by another of 
the Merck literature groups, for coding disease information. This is classi¬ 
fied rather than a nonclassified list; the numbers are not random and can¬ 
not be superimposed to the same extent as random codes. Experience shows, 
though, that several diseases can be superimposed without serious problems 
of false selection. After six months of experience with this technique, it 
appears that it will save searching time for searches involving very specific 
diseases. Indexing time is lengthened because of the need to consult both 
SN 8 and the original dictionary, and because of the intricacies of some 
of the decisions necessary for proper consistency in the use of SN. The 101 
continues to search satisfactorily, in one pass through the machine, the 
combinations of searching terms needed from any or all of the three fields 
used for coding subjects, diseases, or drugs. 

Cards and Card Design 

Mark-sensing cards are used for preparation of the sorting index. Figures 
10-3 and 10-4 show the two sides of the mark-sensing card as it is utilized. 
Perhaps it should be explained that the numerals of a mark-sensing card are 
enlarged and that one side of the card represents only 27 columns of a 
standard sorting card. Therefore, the punches resulting from the pencil 
marks placed on both the front and back of a mark-sensing card, 54 columns 
in all, fit easily into a standard punched card of 80 columns (Figure 10-6). 

After being coded, the mark-sensing cards are run through an electrically 
operated punch that is activated by the graphite on the cards. The mark¬ 
sensing card is punched and used to prepare two duplicates of itself. It is 
then discarded; it cannot be used for searching purposes because the pencil 
marks interfere with the reading of its punches, making sorting inaccurate. 

Two identically punched cards are prepared for each reference so that one 
can be used for sorting and the other can become part of the serially filed 
deck used for maintaining the system. The chief use of the latter is as a 
master deck for punching replacement cards when they are needed. 

Figures 10-3 and 10-4 can be analyzed more closely to demonstrate how 
the coding is applied to the card. There are two random fields (columns 

• American Medical Association. Standard Nomenclature of Diseases and Opera¬ 
tions. 4th ed. Philadelphia, Blakiston, 1952. 



RANDOM CODES FOR LITERATURE CODING 


245 


0001 *>«s\ 


•IIimIIimIimimiIm 

i i I <tt i • • anmnmannnaiia 

lltllltllI I 11 I 111 ill 1 I 

222l222l2222222l222222 
lllssimlissuilnsn 
4 4 4II4 4144444444444444 
llsilssssssisslsmsss 
III III ill'll I ill K <1S • 11 
iliMln/imnmm 
IIiIIiiiiiiiiiiiiiiiii 
I 111 I 11 III 111 I 11111111 


1111111119 

1111111111 

2323233222 

3333333333 

4444444444 

smsmss 

1111111141 

1111111111 

llllllllll 

minim 


iiiiiIIIiiiiiiiii 

1111111 iii 111 1 1 1 1 
22322222222222222 
33333333333333333 
44444444441444444 
S9SSSSSSSSSSSSSSS 
lllllllllllllllll 

ni%iniiiiiiiiii 

llltlllllmlllll 

mmmlilmn 


iiiiiiiiii 

1111111111 
2222222222 
3333333333 
f 444444444 
SSSS9S99SS 
llllllllll 
1111111111 
llllllllll 
lltlllllll 


lllllll 

llllllll 


mm annua 

unit 1 

llllllll 

2222222 

22222222 

3333333 

33313131 

4444444 

44444444 

S9SS39S 

smms 

lllllll 

11111111 

1111111 

11111111 

lllllll 

iiiiiiti 

lllllll 

11111111 




Figure 10-6. Standard IBM punched card with literature reference coded. 


1-10, 31-40) of 10 columns each, devoted respectively to subject words and 
to the names of chemical compounds, as described previously. The names of 
authors are coded in columns 11-14. The first four consonants of the name 
form the code. The last two digits of the year of the reference are punched 
directly into columns 15-16. The journal, or other source for the reference, 
is coded in columns 17-20. Journal codes were devised by assigning four 
digit numbers in spaced sequence to an alphabetical list of the journals 
taken by the library. Example: 

0160 American Chemical Society, Journal. 

0170 American Dental Association, Journal. 

0180 American Documentation 

0190 American Drug Manufacturers' Association, Proceedings. 

Column 41 is labeled “special.” It is used for direct punches to indicate a 
paper authored by a company staff member, a paper about a company prod¬ 
uct, and the color categories of the card that are to be used to record the 
reference. An explanation of the latter will follow. Columns 42-50 are 
used for coding the disease classification explained earlier in this chapter. 
The serial number of the reference (serial numbers assigned as references 
are indexed) is punched in columns 23-27. 

Keeping the sorting deck in any kind of order is unnecessary until the 
volume of cards begins to make the searching time too long. The amount of 
time considered necessary varies from search to search, depending on the 
circumstances. However, in this library the goal is to keep machine time 
for an “ordinary” search to 15 minutes. It was recognized from the be¬ 
ginning that in dealing with journal references it would be useful to divide 
the sorting deck chronologically. This is done, in increments of one year, 
but even one year’s accumulation (15,000 cards) is too much to sort every 
time a question is asked. To obviate this, colored cards have been used for 


















































246 


PUNCHED CARDS 


punching the sorting deck. Four colors are used to denote: (1) human 
clinical papers, (2) veterinary clinical papers, (3) experimental papers in 
the field of biology, (4) all others. The cards are filed in blocks, according 
to color, within each yearly division. The categories that the colors denote 
were chosen as being representative of the types into which most of the 
questions received can be divided. For some questions more than one of the 
color categories needs to be searched. 

Getting The Reference into the System and Out Again 

Persons desirous of knowing how the system operates may be interested 
in the following details: 

Steps in Indexing and Coding a Journal Reference: 

Professional Personnel 

1. The indexer reads the title, the first paragraph and the summary 

of the article, carefully scanning the body of it. 

2. a. Subject words (up to 16) are assigned from the subject diction¬ 

ary to describe the article. 

b. Diseases mentioned are looked up in SN. 8 Their classification 

numbers are written in the box provided on the reference card 
shown in Figure 10-2. 

c. The chemical compounds or trade names of compounds are 

noted for coding in the chemical field (there is an authority 
file kept on a Wheeldex). 

d. The group of terms printed on the bottom of the reference card 

(Figure 10-2) are scanned. All terms that pertain to the article 
are indicated by a check mark. 

Clerical Personnel 

3. The reference card (Figure 10-2) is typed, giving the complete 

reference and the subject tracing for the article. This card bears 
a serial number, under which it is filed. 

4. The code numbers are obtained for the terms indexed and a mark¬ 

sensing card is prepared by making pencil marks over each num¬ 
ber to be punched. These marked cards are punched automati¬ 
cally by a mark-sensing punch. 

5. The punched cards are checked for errors and then filed. 

Steps in Making a Search: 

Professional Personnel 

1. The reference librarian translates the question into the terms of 

the system, choosing the most definitive ones for the search. 
Professional or Clerical Personnel 

2. The necessary code numbers are obtained. 



RANDOM CODES FOR LITERATURE CODING 


247 


3. These codes are set on the selection panel. The machine then de¬ 
livers combinations of the terms being searched, as explained 
previously. The machine can also be used to sequence the selected 
cards by serial number and to print a list of those serial numbers. 

Summary 

This chapter describes an application of random superimposed coding 
for indexing journal references in the fields of medicine, pharmacology, and 
the allied sciences. The machine employed is the IBM Statistical Machine, 
Type 101. 

Random coding is used for indexing chemical compounds and subject 
words describing a reference. In conjunction with this, a classification 
system is used for coding the names of diseases. All the information about 
an article including authors, date, and source is punched into one card. Any 
combination of the “bits” of information punched into a card can be exam¬ 
ined in one pass through the machine. An auxiliary panel that makes it 
possible to set such search codes by means of a dial system is shown by 
photographs. 

The indexing technique is described in detail in an attempt to illustrate 
some of the functional aspects of descriptive indexing. Searching is made 
more productive by the use of an auxiliary panel board w r hich delivers 
the available variations of the information sought into separate pockets 
of the machine. 



Chapter 11 

SEARCHING METALLURGICAL LITERATURE 


Allen Kent and James W. Perry 

Center for Documentation and Communication Research 
Western Reserve University, Cleveland, Ohio 

During 1957, a novel pilot searching service was initiated in the field of 
metallurgy. An experimental literature searching machine is being used to 
scan a file of encoded abstracts in response to questions submitted by vari¬ 
ous industrial and governmental organizations. 

Development of this information service is based on certain processing 
methods and underlying principles which will be discussed under several 
headings as follows: history of the project; codes used; searching equip¬ 
ment; questions, their analysis and programming. 

History of the Project and Introduction 

The problems of coping with the increasing amount and complexity of 
scientific and technical literature which are facing users of metallurgical 
knowledge have long been a source of concern to the American Society for 
Metals. 

The American Society for Metals recognized the need for bibliographic 
control more than a decade ago and took a first step by establishing the 
ASM Review of Metal Literature in 1944. At present, this is an abstracting 
service of the indicative rather than informative type which emphasizes 
the factors of promptness and completeness, without being exhaustive. A 
second step in the ASM program was the compilation and publication of 
the ASM-SLA Classification of Metallurgical Literature in 1950. Although 
the classification system by itself is a tool for organizing literature resources, 
it is specifically designed for use with a hand-sorted punched card system 
(see Chapter 5). 

Both the ASM Review of Metal Literature and the ASM-SLA Classifica¬ 
tion were designed with the needs of the individual metallurgist particularly 
in view. Both services, however, immediately caught the attention of libra¬ 
rians and others who specialize in literature organization and searching, 
and the demand for something still more effective on a larger scale soon 
became evident to ASM. The hand-sorted punched card system is very 
well suited to collections up to about 10,000 documents. To handle the 
much larger collections that are encountered in metallurgical literature, a 
Committee on Mechanized Literature Searching appointed by the Board 


248 



SEARCHING METALLURGICAL LITERATURE 


249 


of Trustees of the American Society for Metals, recommended that ASM 
sponsor a pilot operation to demonstrate the feasibility and advantages of 
applying computing-type equipment to the retrieval and correlation of 
metallurgical literature. 

It was decided by the committee that the need for better methods of 
retrieving and correlating metallurgical literature was urgent, since the 
time was rapidly approaching when it would be cheaper to do a research 
job than to spend the time, effort and money required to do an adequate 
literature search. 

The Center for Documentation and Communication Research, late in 
1955, with a grant of $75,000 from the American Society for Metals, under¬ 
took a five-year program to test and demonstrate the feasibility and useful¬ 
ness of a mechanized searching service to ASM members. To achieve this 
purpose, a pilot-plant operation was required. The basis for such an opera¬ 
tion had been provided during the past ten years both in equipment and 
also in new methods for indexing and coding information preparatory to 
machine searching. 

Highlights of the pilot operation are as follows: 

(1) Approximately 25,000 important metallurgical papers are being 
processed as the basis for pilot plant test and demonstration. 

(2) Encoded “abstracts” are being used as the basis for searching and 
selecting operations. The “abstracts” used are telegraphic in character and 
they are particularly suitable for encoding for machine searching 1 . 

(3) The encoding of the abstracts for the 25,000 published papers is 
being conducted in such a way that a wide range of equipment can be used 
to conduct searching, selecting, and correlating operations. 

(4) The editing to produce telegraphic-style abstracts and their subse¬ 
quent encoding are based on techniques that make explicit for searching 
purposes both the generic significance and the specialized meaning of the 
terminology used in individual abstracts to express important aspects of 

1 See, for example, J. W. Perry, Allen Kent, and M. M. Berry, “Machine Literature 
Searching,” pp 100-108, New York, Interscience, 1956; Allen Kent and J. W. Perry, 
“New Indexing-Abstracting System for Formal Reports, Development and Proof 
Services, Aberdeen Proving Ground,” Am. Doc., 8, No. 1, 34-36 (1957); J. W. Perry 
and Allen Kent, “The New Look in Library Science,” Appl. Mechanics Revs., 9, 
No. 11, 457-60 (1956); Allen Kent and C. R. Flagg, “Abstracting, Coding and Search¬ 
ing the Metallurgical Literature for ASM. The WRU Searching Selector,” in J. H. 
Shera, A. Kent and J. W. Perry, eds., “Information Systems in Documentation,” 
New York, Interscience, 1957; J. W. Perry and Allen Kent, “Tools for Machine Litera¬ 
ture Searching: Semantic Code Dictionary; Applications; Searching Selector,” New 
York, Interscience, (1958); M. R. Hyslop, “Inventory of Methods and Devices 
for Analysis, Storage and Retrieval of Information,” in J. H. Shera, Allen Kent 
and J. W. Perry, eds., “Documentation in Action,” pp. 128-130, New York, 
Reinhold, 1957. 



250 


PUNCHED CARDS 


subject matter. A code dictionary embracing about 10,000 frequently en¬ 
countered scientific and technical terms is available and has been under¬ 
going expansion to include metallurgical terminology. 

(5) Informative abstracts from Metallurgical Abstracts, the Journal of 
the Iron and Steel Institute, and Chemical Abstracts are being used during 
the first two years of the program as the basis for preparing the encoded 
telegraphic-style abstracts for the pilot plant. Starting Sept. 1,1957, original 
publications and papers for a test group are being used as the basis for the 
encoded abstracts (as well as for conventional abstracts). 

(6) A limited searching service is being provided to ASM members at 
present (1958). (This will enable the market potential for the proposed 
service to be evaluated at a relatively early date. It is intended that this 
undertaking shall be placed on a self-supporting basis.) 38 

(7) The “pilot-plant” testing program is planned to extend over a total 
of five years. During the first two years attention was directed to the 
development phase. 

Codes and Methods for Analysis 

Encoding for machine searching requires that metallurgical information 
shall first be analyzed. An analyst reading an article can prepare both an 
abstract in a conventional form ready for publication and also a standard¬ 
ized telegraphic-style abstract ready to be encoded for machine searching*. 
[Incidentally, at the same time that this is being performed, the analyst 
may also indicate what index entries are needed for the conventional 
subject index provided with the ASM Review of Metal Literature 1 .] 

Two aids have been provided the analyst who must perform this task 3 : 

(1) A set of rules has been worked out for preparing standardized tele¬ 
graphic abstracts in such a way as to eliminate the variations and com¬ 
plexities of English sentence structure. 

(2) A series of subject matter analysis forms has been worked out to 
guide the consistent recording of important aspects of subject matter in 
the form of telegraphic abstracts. An example of completed analysis forms 
is given in Figure 11-1 (A-G). The italicized material given at the left-hand 
side of each part of the figure represents the headings presented on the 

* See, for example, Allen Kent and J. W. Perry, “New Abstracting—Indexing 
System for Formal Reports, Development and Proof Services, Aberdeen Proving 
Ground/’ Am. Doc., 7, 36-46 (1957); see also Chapter 6, in J. W. Perry and A. Kent, 
“Tools for Machine Literature Searching,” Interscience, New York, 1958. 

3 Jessica Melton, Manual for Preparation of Telegraphic Abstracts , Western Re¬ 
serve University, Center for Documentation and Communication Research, Cleve¬ 
land, March 25,1957, (Multilithed); see also Chapter 5, in J. W. Perry and Allen Kent , 
“Tools for Machine Literature Searching,” New York, Interscience, 1958. 

** M. R. Hyslop, “Forecast of an Information Center,” Metal Progr., 74, 108-111 
(July 1958). 



SEARCHING METALLURGICAL LITERATURE 


251 


Form A 


Properties given for:) 

Semiconductors, 

Material processed: > 

Binary compounds. 

Starting material: j 

Crystal/single/ n -ty pc 

Component: 

PbS, PbSe, PbTe 

Properties given: 

Semiconductivity 

Form B 

Material processed: 

Welds, Metal, Alloy 

Property influenced: 

Wear/mechanical, Abrasion 

Influenced by: 

Encounter % atoms. Area % contact 

Form C 

Process: 

Brazing/ torch 

By means of: 

Flux 

Condition: 

Vapor 

Form D 

Material processed: 

Containers % bromine 

Component: 

Ni, Monel, Hastelloy, Pb, Steel, Teflon 

Testing technique: 

Immersion, Corrosion 

By means of: 

Bromine 

Condition: 

Wet, Dry 

Property determined: 

Resistivity % corrosion 

Form E 

Product: 

Alloy/N-155 

Component: 

Fe, Cr, Ni, Co 

Properties given: 

Physical 

Form F 

Machine or device: 

Vacuum furnace 

Rating , Size: 

Capacity 1,000 lbs.; Commercial 

Function: 

Melting 

Material processed: 

Steel 

Property influenced: 

Resistivity % temperature/high 

Form G 

Subassembly: 

X-ray unit/Seifert 

Rating , Size: 

Voltage/high 

Focus/fine 

Function: 

Measurement, Radiation 

Material processed: 

Metals 

Fe/gamma 


Figure 11-1. Subject analysis forms for preparation of telegraphic abstracts. 
Italic headings at left-hand side of each part correspond to headings represented on 
the analysis forms; bold face material at right-hand side represents indexing infor¬ 
mation provided by analyst. 

analysis forms; the material given in bold-face type at the right-hand side 
of each part of the figure, opposite the italicized headings, represents the 
index information provided by the analyst. It should be noted here that 
any combination or number of analysis forms may be used and that they 
may be altered as required to record adequately the subject matter of a 



252 


PUNCHED CARDS 


given document. Such alteration must remain, of course, within the pro¬ 
visions of the rules for generating the standardized telegraphic abstracts. 

The next step in the process is to encode the individual terms and phrases 
of the telegraphic-style abstracts. A semantic code dictionary 4 has been 
developed in which codes for specific terms express their meaning in such 
a way that related generic terms are made available as reference points 
for defining and conducting searching operations. For example, the code 
for “steel” will permit searches to be performed, not only to “steel,” but 
also, more generically, to: 

“metal alloys containing iron” 
or to “alloys containing iron” 
or to “ferrous metals” 
or to “metals.” 

Also, the code for “length” will permit searches not only for “length,” 
but also, more generically, to “material properties,” or to “property.” 

The utility of this type of searching possibilities will become increasingly 
evident as the file of encoded abstracts continues to expand. Suffice it to 
say, for now, that sufficient flexibility and capacity are being provided 
for coping with diverse questions and with large files in a fashion not 
feasible with previous methods. 

The code dictionary is maintained in the basic form of punched cards 
so that automatic procedures analogous to machine translation methods 
may be used to convert the telegraphic abstracts into encoded form. 

The next step is to record the encoded abstract on punched paper 
tape—or other appropriate searching media, e.g., Minicards, magnetic 
tape, etc. 

Searching Equipment 

Several different machines can be used to accomplish searching of this 
type of material, as discussed earlier in the chapter. These include digital 
computers, computerlike devices such as the IBM X-794 and devices of 
the type exemplified by the WRU Searching Selector. These machines can 
be either specially programmed or are specifically designed to accomplish 
the type of searching to be described in the latter part of this chapter. 
In addition, the Eastman Kodak Minicard equipment can perform many 
types of searches that are made possible by encoding abstracts along the 
lines indicated above. Relatively minor electronic modifications would 
enable the Minicard equipment to perform the full range of searching 
operations that may be performed with encoded abstracts. 

4 J. W. Perry and Allen Kent, “Tools for Machine Literature Searching,” New 
York, Interscience, 1958. 



SEARCHING METALLURGICAL LITERATURE 


253 


Digital computing equipment is, of course, available from the major 
business machine companies such as Sperry Rand (the series of machines 
known as Univac), International Business Machines (the “700” series), 
and others. The Minicard equipment referred to has been developed on 
Air Force funds by the Eastman Kodak Company in cooperation with the 
Magnavox Company of Indianapolis, Indiana. Announcement of the 
eventual availability and marketing of Minicard equipment was made as 
early as October 1955, in the Wall Street Journal and in other periodicals. 
The IBM X-794 was a commercial development that, it appears, could 
be made available within a reasonably short time. 

The WRU Searching Selector was designed to search punched tape 
and specifically to perform the logical operations required for the effective 
searching of encoded abstracts. The WRU Searching Selector is char¬ 
acterized by very simple circuits, by unusual capabilities for performing 
up to ten simultaneous searches based on complex logical relationships, 
and by relatively low speed of operation. 6 

The WRU Searching Selector is able to perform the following functions*: 

(1) Use patterns of holes in punched paper tape to record sequences of 
symbols. In this way, the characteristics of documents may be recorded 
one after another for subsequent search by the selector. (Individual symbols 
and combinations of symbols may be used to record the characteristics of 
documents in the same way that individual letters and combinations of 
letters are used to denote words in ordinary writing. It should also be 
noted that meaning may be ascribed to any combination of symbols as 
may be appropriate.) 

(2) Read the tape by means of a Flexowriter and convert the patterns 
used to record successive symbols into corresponding electrical pulses, 
which then activate the discriminating unit. 

(3) Detect those characteristics and combinations of characteristics 
which typify the subject contents of documents that are of pertinent in¬ 
terest. The discriminating unit is conditioned to detect such characteristics 
by appropriate wiring of the plug board prior to initiating a given search. 

(4) Type out automatically the serial numbers of those documents whose 
characteristics correspond to the requirements of a given search. The 
scope of a search is expressed by specifying that the documents of pertinent 
interest shall have some one characteristic or some combination of char- 

* Design work is now under way for a high-speed counterpart of this searching 
selector. The scanning rate will be 200,000 to 300,000 abstracts per hour with 20 
searches performed simultaneously. 

* Exerpted from J. W. Perry and Allen Kent, “The New Look in Library Science,” 
Appl. Mechanics Rev., 9, No. 11, 457-460 (November 1956). See also Chapter 18, in 
J. W. Perry and Allen Kent “Tools for Machine Literature Searching,” New York, 
Interscience, 1958. 



254 


PUNCHED CARDS 


acteristics. Possibilities for specifying combinations of characteristics are 
outlined below. 

The sequences of symbols may be organized into combinations analogous 
to “syllables” from which “words” may then be built up and from which, 
in turn, combinations analogous to “phrases,” “sentences,” and “para¬ 
graphs” may be built up. If the capital letters, A, B, C, D, etc., are used 
to designate individual symbols, then letters with subscripts may be used 
to designate various levels of combinations as follows: 

Ai, Bi , Ci, Di, etc., for “syllables” 

A 2 , B 2 , C 2 , D 2 , etc., for “words” 

A 3 , B 3 , C 3 , D 3 , etc., for “phrases” 

A 4 , B< , C 4 , D<, etc., for “sentences” 

A 6 , B 6 , C 6 , D 6 , etc., for “paragraphs” 

A*, B«, C«, D«, etc., for “messages”. 

This ability to organize characteristics into sets analogous to “phrases,” 
“sentences,” etc., is important in preventing false association of char¬ 
acteristics when searching. For example, by proper “phrasing” it is pos¬ 
sible to prevent the properties of one alloy being incorrectly attributed to 
some other alloy. 

At any level, which we may term the “n-th” level, each combination 
consists in general of a number of component combinations at the “n — 1 ” 
level. Each of several “n-th” level combinations, denoted by A„ , B n , 
C„ , D n , etc., may be specified in terms of component units designated by 
A„_i, B n _i, C„_i, etc. Thus, in conducting a search, it may be specified 
as a condition that a document will be identified as being of pertinent 
interest, that at least one “n-th” level combination shall be characterized 
by certain component units. Specification of the component units may be 
set up on the basis of the following relationships. It may be specified 
that: 

( 1 ) All of several components units must be present. This requirement 
constitutes a logical product that may be symbolized, for example, hv: 

A„_i • B„_i • C n -i, etc. 

In specifying logical products, further requirements as to order may be 
imposed. Thus, for example, it may be required that all components 
specified by a logical product shall occur in sequence. For example, it may 
be required that A n -i shall be followed by B n _i and it, in turn, by C„_i. 
This requirement may be symbolized by 


(A n _i • B„_i • C„_i) 

The reverse order of these three components might also be specified as 



SEARCHING METALLURGICAL LITERATURE 


255 


denoted by 

(C„_i • B„_i • A n _i) 

(2) Any one of several component units or, alternately, one or more of 
several component units must be present. This requirement constitutes a 
logical sum that may be symbolized, for example, by, 

A„_j + B„_ x + C„—i, etc. 

(3) At least one component unit must be present but at least one other 
component unit must be absent. This requirement constitutes a logical dif¬ 
ference that may be symbolized, for example, by 

A n —1 B n —1 

Here, also, order may be designated. Thus it may be specified that B„_i 
may not follow A n _i. This requirement would be symbolized by 

(A n —1 B n -l) 

Alternately, it might be specified that B n _i may not precede A n -i and this 
would be symbolized by 

( — B„_i-A„_i) 

(4) Combinatims of component units expressed by complex logical re¬ 
lationships must be present. Such logical relationships as the following may 
be specified 

(A n —1' B n _1 — Cn—l)(Dn—1 + En_l) 

(A n -1 + B„_i)(C n -l — D n -l) E n -1 

(A n -i — B n -i) (C„_i • D„_i — E n -i) 

Any such complex logical relationship may be set up as required at any 
level. Such complex logical relationships may also involve specification of 
sequential order. Using the symbols ( ) to denote order as before, we 
might specify such search requirements as 

((A n —1' B n _1 — C n -l))(D n _i + E n -l) 

(A n _1 + B»_i)(«C„_i - D„-i»E n -i> 

(((A„_! - Bn-l»(Cn-rDn-l - E„_,)) 

Application of these capabilities means that “syllables” may be specified 
in terms of component symbols, e.g., letters, “words” may be specified in 
terms of “syllables,” “phrases” in terms of “words,” “sentences” in terms 
of “phrases,” “paragraphs” in terms of “sentences,” and “messages” 
in terms of “paragraphs.” 



256 


PUNCHED CARDS 


As stated above, the abstract formulation of higher order characteristics 
in terms of their lower order components has been restricted to the special 
case that the “n-th” order characteristic, e.g., a “sentence” shall be speci¬ 
fied in terms of its components at the next lower “n-1” level, e.g., at the 
“phrase” level. The WRU Searching Selector may be readily programmed 
so that a higher order combination is specified in terms of any desired 
combination of lower order characteristics provided only that their order 
is less than “n”. Thus, for example, “sentences” may be specified not only 
in terms of logically defined combinations whose component elements may 
be “phrases” and also “words”, “syllables” and individual symbols. In 
abstract formulation, a characteristic of “n-th” order may be specified: 

(1) as a logical product e.g., AyB*-Cj 

(2) as a logical sum, e.g., Ay + B* + Cj 

(3) as a logical difference, e.g., Ay — B* 

(4) as a complex logical combination e.g., 

l(A*-B t ) - C H ] [Dy + EJ (A*• B„) (Cy - D.) E* [A, - By] [(C*-D*) - E,] 
where “e,f, g, h,j, k, l” are each less than “w”. 

In such combinations, two or more component elements of lower order 
than “n” may be of the same lesser order. 

These capabilities enable encoded telegraphic abstracts to be searched 
conveniently and effectively. Various other machines with somewhat 
lesser capabilities for performing the searching operations required are 
available commercially from one or another business machine manu¬ 
facturer, and are too numerous to enumerate here. For example, the 
precursor of the IBM 101 statistical machine—the Census 100—has some 
of the features required for searching encoded abstracts as prepared for 
the American Society for Metals. It must be noted, however, that with 
such machines modification of the encoded abstracts would probably be 
advisable to match the limited capabilities for literature searching by 
these machines. 

Questions, Their Analysis and Programming 

The question chosen as an example for detailed consideration is the 
following: 

“How does the presence of vanadium in titanium alloys affect their cast¬ 
ing?” This simple question permits some of the most important capabilities 
of searching system to be illustrated. In particular, the searching strategy 
may be varied depending on whether it is desired to extend the range of 
selected papers to include those that contain information that may be of 
less direct interest. 



SEARCHING METALLURGICAL LITERATURE 


257 


It is perhaps obvious that reports of experiments and tests directed to 
the casting of vanadium containing titanium alloys or to studies on their 
castability properties will be of prime pertinency to our example question. 
Accordingly, a sharply focused interpretation of our question might be 
formulated, with designation of code elements, as follows: 

“Select those encoded abstracts that mention an alloy (LALL.001) whose 
principal component (KUJ) is titanium (MATL.ll.dTQI) with a lesser 
component (KUJ), vanadium (MATL.ll.DV), when the alloy is either 
the material processed (KEJ) by the process (KAM) casting (CUNS. 025) 
or related terms (CUNS.25X) or when a property given for (KOV) the 
alloy is castability (CUNS.25X.PAPR.004) which is designated either as a 
property given (KWV) or as a property influenced (KAP).” 

Conversion of this statement of the example question into a searching 
machine program requires specification both of logical and also of sequen¬ 
tial relationships between the above cited code elements, such as LALL.001, 
KUJ, etc. Such relationships may be expressed symbolically as follows: 


Code elements (Syllable level) 

Gi = nv. 

Hi = KEJ. 


Ai = LALL. 

Bi = 001. 

Ci = KUJ. 
D! = MATL. 
E! = 11. 

F! = DTQI. 


I, = KAM. 
Ji = CUNS. 
K, = 025. 

L x = 25X. 


Mi = KOV. 
Ni = PAPR. 
Oi = 004. 

Pi = KWV. 
Qi = KAP. 


(Here the three letter codes with K as initial letter are role indicators which 
indicate mode of involvement of the term whose code follows within a sub- 
phrase.) 

Subphrase level 

A t = (Hi-Ai-Bi) Ct = (Ci-Di-Ei-Fi) 

Bj = (Mi*Ai*Bi) Dj = (Ci-Di-Ei-Gi) 

E, = (Ii-Ji (Ki + Li)> 

Ft = ((Pi + Qi)«Ji-Li-Ni-Oi») 

Phrase level 

At — (A2 • C2 ■ Dj) C3 = E2 
B3 = (B2 • C2 * O2) D3 = F2 

Sentence level 
A 4 = (Aa- C3) -f- (B3D3) 

In this encoding of the search, particular importance attaches to the 
combination of code elements denoted by E 2 = (I 1 J 1 (K t + L x )) that is 



258 


PUNCHED CARDS 


to the combination of code elements KAM, CUNS and either 025 or 25X 
when detected within a subphrase in that order. Here the role indicator 
KAM indicates that the term whose code follows in the same subphrase 
denotes a process. The combination of the code element CUNS with either 
025 or 25X occurs in the codes for casting and various related terms 
denoting various specific casting processes, materials used in casting and 
related terms as may be evident by the following examples from the code 
dictionary, 

continuous casting CUNS.25X.CYNT.10X.MWTL.PASS.001 

Junghaus-Rossi process CUNS.25X.CYNT.10X.MWTL.PASS.002. 
Asarco process CUNS.25X.CYNT.10X.MWTL.PASS.005. 

core oil CUNS.25X.FATT.3X.MWPR.24X.MWTL.001. 

foundry CUNS.25X.LACN.001. 

shelf molding CUNS.25X.MWTL.PASS.003. 

founding CUNS.25X.MWTL.PASS.009. 

as well as numerous additional terms for casting processes. 

Thus specification of the combination of code elements CUNS. and either 
025 or 25X is equivalent to citing a lengthy list of single terms relating to 
castings, while the requirement that KAM shall be found with CUNS and 
either 025 or 25X, in that order, effectively selects out that those terms that 
designate casting operations. 

The combinations of code elements designated by C* and D 2 , namely 
MATL. and 11., indicates a class of metals in the ASM-SLA classification 
while DTQI and □ V are special codes for the chemical elements titanium 
and vanadium. Note that specification of titanium as the principal com¬ 
ponent of the alloy is made possible by encoding alloys that the main me¬ 
tallic component is cited first among the components of an alloy. 

The above presented formulation of our example question illustrates its 
conversion to a sharply focused machine selection program to identify 
those encoded abstracts that are virtually certain to be of direct pertinent 
interest. The scope of search may be readily extended, by various altera¬ 
tions in the machine searching program, to accomplish identification of 
additional encoded abstracts that may be expected to refer to information 
of less directly relevant interest. For example, the search might be extended 
to include: 

“As products, castings, and the like comprising titanium alloys contain¬ 
ing vanadium.” (Here the combination of code elements, KWJ and C-NS, 
with either 025 or 25X within a single subphrase will characterize castings 
and similar products when mentioned as products.) 

If Ri is used to designate KWJ, the role indicator for “Product” and Si 



SEARCHING METALLURGICAL LITERATURE 


259 


is used to designate C-NS, then this extension of the search may be sym¬ 
bolized as follows: 

Subprhase level 
G 2 = (Ri-Si (Ki + L0> 

H 2 = <Ci-Ai-B,> 

Phrase level 

E 3 = <G 2 H 2 C 2 D 2 ) F 3 = G 2 

The over-all scope of the extended search (denoted by B 4 ) may then be 
abstractly specified as follows: 

Sentence level 

B< = A 3 • C 3 + B 3 *D 3 + E 3 + A 3 • f 3 

A further extension of the scope of search may be made to include en¬ 
coded abstracts which cite the casting or castability or castings (as products) 
of alloys mentioned by trade name only in the original publication. If such 
alloys are known to contain titanium as principal component and vanadium 
as lesser component, this will be indicated by their codes and the latter, in 
turn, enable the searching selector to be programmed to detect them by 
specifying a suitably defined sequence of code elements, namely; LALL 
and DTQIQ V. In this way, the scope of search may be extended, if desired, 
to include papers which referred to the casting or castability of alloys whose 
trade names provided the only indication that they were titanium alloys 
containing vanadium. Thus, the previously extended search, denoted by 
B 4 may be further extended by setting up additional search requirements 
at various levels as follows: 

Code elements (Syllable level) 

Ti = □TQIDV 
Suhphrase level 

U = (Hi-Ai-Ti) J 2 = (Mr Ai-Tj) 

K 2 = (CrAi-Ti) 

Phrase level 

G 3 = I 2 H 3 = J 2 I 3 = (G 2 K 2 ) 

The overall scope of this further extended search (denoted by C 4 ) may 
then be abstractly specified as follows: 

Sentence level 

C 4 = [(C 3 + F 3 ) (A, + G 3 )] + [D 3 (B 3 + H 3 )l + E 3 + I 3 



260 


PUNCHED CARDS 


It should be understood that both the narrow and the broad interpreta¬ 
tions of our example search, as represented by the logical formulations, A 4 , 
B 4 and C 4 above, may be searched simultaneously. The WRU Searching 
Selector is designed to search ten independent questions at once. 

Result of Search 

The basic operation of the WRU Searching Selector is the scanning of 
a continuous tape in which the punching of successive combinations of 
holes records the succession of symbols that spell out encoded abstracts. 
The scanning operation causes the search requirements to be matched with 
the encoded characteristics of the various abstracts. 

As mentioned earlier, ten searches may be performed simultaneously. 
The machine automatically types out: (1) the serial number of each selected 
article; (2) the number of the search that has been satisfied; and (3) a 
bibliographic citation. For example, the machine may type out: 

7325 1 G. H. Schippereit, R. M. Lang and J. C. Kura. American 

Foundry men’s Society, Transactions, 65, 499-512 (May, 1957). 

to indicate that article 7325 in the file satisfies search number 1 of ten 
searches and that this article is located in American Foundry men’s Society, 
Transactions, Volume 65, May, 1957, pages 499-512. 

After identification, articles are removed clerically from the file and 
presented to the expert analyst for review and evaluation. If the Minicard 
Selector is used instead, the cards selected can contain a microphoto of the 
original paper which may be viewed in a suitable reader. 



Chapter 12 

CLASSIFICATION, SEARCHING AND 
MECHANIZATION IN THE U.S. 
PATENT OFFICE 


B. E. Lanham and J. Leibowitz 
U. S. Patent Office, Washington, D. C. 


Introduction 

This chapter includes references and descriptions of Patent Office fea¬ 
tures such as history, conventional patent searching and classification, as 
well as progress and experiments in mechanized searching. Its purpose is 
to provide the reader with a general overall view of the Patent Office and 
its functions and objectives. For those who desire more specific information 
the listed references may be of assistance. 

Considerable interest has been manifested in the various operations of 
the Patent Office and its experimentation in patent search mechanization, 
and it is hoped that the included descriptive matter will be of value. 

Historical Background of Patent Searching and Classification 

On April 10, 1790, President Washington signed the Congressional Bill 
under provision of Article 1, Section 8, of the Constitution, authorizing the 
grant of patents by the U. S. Government. 

The 1790 Act required as a condition precedent to the grant of a patent 
that satisfactory evidence of novelty, utility, and invention be established, 
which requirements are in existence at the present time. A “prior art search” 
was thus necessary, and since it was apparently limited to the relatively 
few patents issued by American Colonies and States as well as among 
books on mechanics and industrial arts, no need for classification of the 
searchable material was then necessary. 

The first U. S. patent was issued on July 31, 1790, and the total was 57 
on February 21,1793, when a new Patent Act replaced the earlier one. The 
new Act substituted a “registration” system for the “examination” system, 
and that unfortunate replacement continued until the Act of July 4, 1836, 
was passed. At that date 9,957 patents had been issued and the new Act 
reestablished the examination system, including prior art searching—for 
that which had been invented or used before. A patent classification was 
formed, including 22 classes with no subclasses, the search material being 


261 



262 


PUNCHED CARDS 


the manuscripts filed by applicants. The Patent Office was established as 
a distinct bureau with the appointment of the first Commissioner of Pa¬ 
tents. Patent No. 1 of the present series started with the 1836 Act and was 
issued on July 13, 1836, and in 1866 the printing of patent copies was per¬ 
manently started. At the end of 1868 slightly more than 80,000 patents 
had been issued; these were divided into 36 classes in alphabetical order 
of their titles, with some classes containing sections not in accord with 
present subclasses. 

In 1872 the previous alphabetical classification was revised and the is¬ 
sued 131,000 patents distributed among its 145 classes. In 1880 publication 
of the first classification with both classes and subclasses occurred, with 
minor subclasses indented under major ones. None of these early classifi¬ 
cation systems, however, were based on the principles governing the allow¬ 
ance of patents, but they should have been devised on this basis. 

The Classification Division (now Classification Group, consisting of five 
Divisions and a Service Branch) was established in 1898 by authorization 
of Congress 1 • 2 . 

Patentability Requirements and Uses of Classification 

In performance of its function the Patent Office examines applications 
to determine whether or not the applicants are entitled to patents under 
the law. In the language of the Statute, “Whoever invents or discovers 
any new and useful process, machine, manufacture, or composition of 
matter, or any new and useful improvement thereof, may obtain a patent 
therefor, subject to the conditions and requirements of ...” patentability 
as expressed in other sections of the Statute*. The purpose of the search is 
to determine whether the subject matter for which a patent is sought 
satisfies the statutory requirements for novelty and invention. 

No quantitative yardstick for invention has been devised. Certain criteria 
have been developed, however, as a result of years of experience and in 
view of various appellate and judicial decisions which aid in determination 
of the question of invention. As a broad statement as to these factors, if 
the subject matter sought to be patented involves an obvious dissimilarity 
to the most nearly similar thing known, it is not regarded as inventive, and 
the same applies to advances which would be considered obvious to a person 
having ordinary skill in the art. The Examiner is thus interested not only 
in identical but in all related and analogous subject matter. 

1 “The Story of the United States Patent Office, 1790-1956,” third edition. Super¬ 
intendent of Documents, 25 cents. 

* M. F. Bailey, “History of Classification of Patents,” J.P.O.S., 18, 463-507, 537- 
575 (July and August, 1946). (Reprints are available from Research and Development, 
U. S. Patent Office, Washington 25, D. C.) 

3 “Patent Laws.” Superintendent of Documents, 25 cents. 



CLASSIFICATION, SEARCHING AND MECHANIZATION 


263 


In addition to the searches made by the Examiner there are related 
types made by others. Before filing an application, an inventor usually 
makes a “pre-ex” search to determine the probable novelty of his invention. 
Those who question the validity of a patent and wish to prove in court 
that it was erroneously issued, make an exhaustive search in an attempt to 
find anticipatory references against the claims. Others are interested in a 
general study of patents to determine developments in certain fields of 
endeavor, known as “state of the art” search. The claims of unexpired 
patents are studied to determine if manufacture of a certain product or 
performance of a certain process may or may not result in an infringement 
suit. Research scientists find that patent literature may furnish valuable 
background material for a project and thus by finding what had already 
been achieved, avoid duplication of effort. 

The Classification System 

The Patent Office classification system is intended to provide facilities 
for storage and location of patents which relate to all branches of science 
and technology whereby searchers may, within a reasonable time, have 
available the art segments of interest. This purpose has been appreciably 
but not entirely accomplished. 

At present there are approximately 309 classes which contain a total of 
over 52,000 subclasses. The number of U. S. patents issued is over 2,800,000 
(this figure does not include reissue, design or plant patents). The class 
and subclass schedules are contained in the Manual of Classification 4 which 
includes an alphabetical index of titles with reference to pertinent classes 
and subclasses. Revision of all the classes has been practically completed 
under modem methods, and all revised ones, including their subclasses, 
have definitions and notes as to content and scope of their subject matter 
and relationships, differences, and search suggestions as to other pertinent 
ones 5 . 

The methods of classifying patents are quite complex and are sometimes 
considered inconsistent by inexperienced classifiers and searchers. The 
following examples illustrate a few general types. 

All organic and inorganic chemical compounds, regardless of their dis¬ 
closed utility, are classified on the basis of their chemical constitution. 
Most compositions of matter, i.e., mixtures of two or more ingredients, are 
classified primarily on the basis of their necessary functions or inherent 
properties rather than upon the basis of ingredients. Such primary groups 

4 “Manual of Classification,” plus Alphabetical Index. Superintendent of Docu¬ 
ments, $12.00. 

* “Definition Bulletins.” Purchasable from Patent Office. Identification number 
and price of Bulletin obtainable upon receipt of class number. Ex.: Class 260, Bulle¬ 
tin No. 200, 80 cents. 



264 


PUNCHED CARDS 


are subdivided into subclasses on the secondary basis of selected ingredients. 
Processes of making compounds or compositions are classified, in most 
instances, on the basis of the resulting products. Processes, such as manu¬ 
facturing (Class 29, Metal Working), the application of coating material 
to a base (Class 117, Coating, Processes and Miscellaneous Products), are 
classified on the basis of their function or ultimate result, and the charac¬ 
teristics of the subclasses are the selected operations included in the proc¬ 
esses. Processes which do not result in a product (Class 209, Classifying, 
Separating and Assorting Solids) are classified on the same basis. In some 
classes, such as 209, the processes and apparatus for performing the same 
functions are classified together since the common search is consistent and 
coextensive. 

Manufactured products are usually classified according to their disclosed 
and necessary function or utility (Class 2, Apparel; Class 81, Tools). 

Machines are generally classified on the necessary mode of operation and 
effect produced rather than upon the specific material handled (Class 202, 
Distillation; Class 241, Solid Material Comminution or Disintegration). 
However a few types of machines are classified on the basis of the material 
handled (Class 80, Metal Rolling). 

Class 241 is an illustration of a modern machine class. It contains three 
main sections. The first includes subject matter (the class machine) com¬ 
bined with features such as means to prevent explosions therein which are 
not necessary for the essential functioning of the machine. The second 
includes the machine per se, and the third subcombinations of the machine, 
such as disc grinding elements, which are not classified elsewhere. Other 
parts of Class 241 machines, such as motors, alloys, etc., not included in 
the class, involve other search fields. 

As a general statement, a patent is classified on the basis of the inventive 
or claimed subject matter, and since consideration of patentability of 
claimed subject matter in an application is not limited to what has previ¬ 
ously been patented, the major amount of unclaimed disclosures in a patent, 
as well as in other types of literature, are of searchable value. Thus cross- 
references of patents based on such disclosures are placed in pertinent 
classes and their subclasses. 

More specific details of the foregoing references to general illustrative 
classification have been previously published 2, M . 

• “The Classification of Patents” (2d Revision). (Copies or reprints available from 
Research and Development, U. S. Patent Office, Washington 25, D. C. Copies of the 
first edition only are now available.) 

7 M. C. Rosa, “Problems of Classifying Chemical Patents,” J.P.O.S., 19. 241-261 
(April, 1947). 

• B. E. Lanham, “Chemical Patent Searches,” Ind. Eng. Chem., 43. 2494-2496, 
November, 1951; J.P.O.S., 34, 315-323 (May, 1952). 

• Bailey, M. F., B. E. Lanham and J. Leibowitz, “Problems of Classification and 



CLASSIFICATION, SEARCHING AND MECHANIZATION 


265 


The classification is based upon the criteria of patentability, and the 
following excerpt from the Manual of Classification illustrates its principles: 

“As all patentable arts or instruments are created for an ulterior utility, 
the characteristic selected as the basis of classification is that of essential 
function or effect. Arts or instruments having like functions, producing like 
products, or achieving like effects are brought together; but the functions 
or effects that serve as a basis of classification must be proximate or essen¬ 
tial, not remote or accidental.” 

The foregoing class examples illustrate what appear to be inconsistencies 
in establishment of classes. However the arrangement of various classes 
as related to different subjects is made on the basis of the ultimate property 
or utility to be searched as specified in the above quotation. Thus the in¬ 
consistencies are more apparent than real. 

Conventional Patent Searching 

The rather broad patent searching operations described here are intended 
as examples and suggestions to benefit those whose types of searches have 
been mentioned 7 - ®- 1<Ml . 

The problem in patent searching is not particularly different from others 
where it is desirable to isolate certain information from a vast and hetero¬ 
geneous field of subject matter. As to questions of patentability, specific 
differences exist as to the types of relationships sought, the comprehen¬ 
siveness of the subject matter and the variability in search requirements. 

Prior to the start of a search, whether it is to be manual or mechanized, 
a thorough study of the subject matter of interest, as well as the purpose 
of the search, should be made. All aspects should be analyzed and verified 
and determination made as to whether the search is to be limited to the 
precise product, process or apparatus or to equivalents thereof, or to generic 
or specific variations or fragments thereof. 

If the search is promoted by a desire to file an application for a patent, 
the Examiner’s search viewpoints and patentability requirements should 
be reviewed. Fragments of disclosures of the claimed invention may be 

Documentation in the United States Patent Office in the Field of Petroleum and 
Allied Subjects,” Third World Petroleum Congress, The Hague, 1951. Proceedings, 
Section X. P. 13-21. 

10 S. M. Newman, “Problems in Mechanizing the Search in Examining Patent 
Applications,” (Copies are available from Research and Development, U. S. Patent 
Office, Washington 25, D. C.). 

11 Lanham, B. E., J. Leibowitz and H. R. Roller, “Advances in Mechanization of 
Patent Searching—Chemical Field.” April 11, 1956. J P.O.S., 38,820-838 (December, 
1956). (Copies are available from Research and Development, U. S. Patent Office, 
Washington 25, D. C.) 

** H. F. Lindenmeyer, “What does the Patent Office Scientific Library Have to 
Offer the Chemist?,” J.P.O.S., 36, 463-481 (July, 1954). (Copies are available from 
Research and Development, U. S. Patent Office, Washington 25, D. C.) 




266 


PUNCHED CARDS 


combined in separate documents as anticipatory, provided such type of 
combination is within the suggestion of one or more documents. 

After the details concerning the scope and approach have been deter¬ 
mined, they should be kept in mind continuously throughout the search 
while studying each document. 

Unfortunately, no infallible or obvious procedure is always available to 
obtain prompt or ultimate identification of the specific subclass for each 
item to be searched. The complete search is not often limited to one or a 
group of subclasses or even to a single class, and it may thus be necessary 
to identify additional search fields for various phases of the subject matter 
sought. 

In many cases, use of the alphabetical index of the Manual of Classifi¬ 
cation will help to locate the proper search areas quickly and accurately. 
If the term sought is not found, its synonyms should also be investigated. 
If the index is not used then the titles of the main classes should be scanned 
to select the one that appears pertinent. The classes in the Manual are 
given in numerical rather than subject order. The class titles usually indi¬ 
cate their relationship to the required search, but if there is any doubt other 
related classes should be compared. The definitions and notes of a class 
and its subclasses usually verify the search field. 

After selecting the pertinent class, the titles of the major subclass group¬ 
ings which appear in the first line of indentation should be read in sequential 
order. Once a title has been located which identifies the subject matter 
the coordinate major subclasses under it can usually be ignored. Minor 
and sub-minor subclasses indented under the selected major subclass should 
be investigated to determine if their titles also relate to the search required. 
Indented subclasses usually include species, their major subclass is generic, 
and includes species not provided for in its indentations. Almost every 
class includes a subclass bearing the heading “Miscellaneous”; this is a 
pigeon hole for patents which fall under the class definition, for which 
there is no specific subclass. Varying scopes of a given subject matter are 
not separately classified. The subclass numbers in many classes are not in 
numerical order and serve only for identification rather than superiority. 

An example of the major and minor subclasses is illustrated below. 

Class 260, Chemistry, Carbon Compounds, certain subclasses being listed 
in the following order: 

239. Heterocyclic Carbon Compounds 

298. Azoles 

302. Thiazoles 

303. Anthrone or anthraquinone nuclei 

304. Arylenethiazoles 

305. 2-amino 

306 . 2-thio 



CLASSIFICATION, SEARCHING AND MECHANIZATION 


267 


The full title of subclass 306, Class 260, is as follows: Chemistry, Carbon 
Compounds, Heterocyclic Carbon Compounds, Azoles, Thiazoles, Arylene- 
thiazoles, 2-thio. An attempt to locate the proper search area should defi¬ 
nitely follow class plans and structures as well as definitions and notes, 
as created by various classifiers, since these factors differ among classes. 
If any difficulty or doubtfulness should arise in determining a proper field 
of search, the Patent Office will supply available information as to the 
identity of the pertinent classes and subclasses, provided the request in¬ 
cludes specific details of the subject matter sought 13 - u . 

Classification System Inadequacies 

The shortcomings of Patent Office and other classification systems are 
to some extent similar 16 . The effectiveness of a system is dependent upon 
how closely the basis of its establishment is correlated with the basis of 
the required searches, but it is impossible for the patent classifier to provide 
for or even anticipate all the search viewpoints to be desired within the 
scope of a given class. Such problems detract from the efficiency of the 
searches made by the Examiner as well as by others. A few examples of 
such difficulties will be illustrated with respect to the chemical field. 

When confronted with a search for a specific chemical compound, wherein 
all of the structural characteristics are set forth, the structural group present 
in the formula which appears highest in the subclass schedules will identify 
the precise field of search. Thiazole in Class 260 is an example. Generic 
searches, however, present a major problem. For example, if it should be 
desired to find disclosures of all compounds which contain an azole structure 
regardless of any other structures which may be attached thereto, the 
“azole” category in the schedule would not provide an adequate search. All 
superior categories may contain disclosures of the type sought but they 
have been classified on the basis of other fragments. 

The search for generic processes involves the same problems. Those which 
result in a specific compound are classified therewith, but when the proce¬ 
dures are not limited to specific reactants and product the practical search 
field is not ordinarily identifiable. 

Certain types of composition searches are also extremely difficult. Where 

u “Information Concerning Patent Classification and Patent Records.” Two page 
circular. (Copies are available from Research and Development, U. S. Patent Office, 
Washington 26, D. C.) 

14 B. E. Lanham, “Services Available from the Patent Office,” Special Libraries, 
46, 25-28 (January, 1955). Elaboration on (13). (Copies are available from Research 
and Development, U. S. Patent Office, Washington 25, D. C.) 

u D. D. Andrews, “Modernizing Chemical Patent Classification,” Presented at 
128th ACS Meeting, Minneapolis, September 16,1955. (Copies are available from Re¬ 
search and Development, U. S. Patent Office, Washington 25, D. C.) 



268 


PUNCHED CARDS 


the search is to determine if compound X has been disclosed as an ingredient 
in lubricants, and such compound is not included in the title or definition 
of an ingredient subclass, all patents in the 82 lubricant subclasses in Class 
252, Compositions, must be investigated. 

Another composition search problem is quite important, both to the 
Patent Office and inventors. A specific composition disclosed only for use 
as a detergent is not patentable over a previous disclosure of the same com¬ 
position for another use, such as an adsorbent. The complete search for 
such a detergent composition, then, would possibly require investigation 
of all composition disclosures distributed among numerous classes. 

No classification is based on adhesive compositions, and without a knowl¬ 
edge of the specific ingredients, an almost unlimited search in sections of 
several classes is required. 

Another frequent approach to searching is by one who wishes to learn 
all of the uses of a material, such as titanium in alloys, compounds, manu¬ 
factured articles, etc. No such search can be indicated since the desired 
information may be found in many of the classes. 

Such difficulties cannot be avoided in the present classification system, 
nor is it practical from various viewpoints to revise and enlarge the system 
to a sufficient extent. 

In 1946 the Patent Office considered the development of mechanized 
search procedures to provide facilities for searching any given subject 
matter from any required viewpoint. 

The early experiments involved studies with respect to the use of edge- 
notched cards and the “unit card system.” Since such methods were even¬ 
tually determined to be impractical for Patent Office objectives they were 
abandoned in favor of what was considered a more feasible approach, the 
general description of which will be set forth later in this Chapter. The 
first will relate to the experiment which culminated in machine searching 
tests in the Spring of 1950; the second will deal with the current program 
which started subsequent to the recommendations of the Bush Committee 1 *. 

THE FIRST EXPERIMENT 

Subject Matter 

The first experiment was in the field of medicinal compositions, and a 
sample group consisting of 441 patents was selected to constitute the sub¬ 
ject matter. The disclosures of these patents were considered to consist of 
two basic types of information units—ingredients and functions. The 
ingredients were chemical compounds, free elements and so-called “com¬ 
plexes,” a term applied to materials described by language other than or 

14 “Report to the Secretary of Commerce by the Advisory Committee on Applica¬ 
tion of Machines to Patent Office Operations,” Washington, D. C., December 22,1954. 



CLASSIFICATION, SEARCHING AND MECHANIZATION 


269 


in addition to the terminology of chemical compounds. Complexes included 
such things as “poppy seed oil,” liver “extract,” milk, etc. 

The term “function” related to nontangible disclosures such as proper¬ 
ties, uses, and behavior of materials. Such terms as “hormone,” “lubricant,” 
“malaria” were considered to be functions. 

The Problem 

A patent search involves, in effect, a question and an answer. The ques¬ 
tion may be formulated as follows. Is there a disclosure anywhere of the 
concept expressed in the claimed subject matter? The answer is provided 
by either finding or not finding such disclosure. 

Several features are significant in patent searching. First, the searcher 
is initially unaware of the existence or non-existence of a disclosure; in 
contradistinction to a search where there is known to be a positive answer, 
such as a quest for the date of an historical event. Since a presumption 
that there is no reference to the disclosure is made on the basis of failure 
to find any, it is important that no pertinent detail, or place where such 
detail may be located, be overlooked. 

Second, the search is based not on words but on the meaning or import 
of the words. Pertinency of disclosure relative to the subject matter is 
evaluated on the basis of the meaning of the disclosure as compared with 
the meaning of the subject matter of the claims. 

Third, the search is often “generic,” not only because claims may be 
presented in generic form but also because the Examiner may be searching 
for “related” subject matter to determine the question of invention in 
addition to the question of novelty. A generic search is met by finding any 
specific embodiment within the scope of the genus. It is not practically 
possible, however, for the searcher to envision all the members embodied 
by the genus and to thus express his search in terms of a collection of these 
specific members. 

It will be evident that the search question and the disclosure which an¬ 
swers it will not ordinarily be in the same language or context. The subject 
matter of the search will very often be included within a more comprehen¬ 
sive disclosure context. Thus A -J- B is included within A + B + C and 
A B is included with A B C; a search for a disclosure of mixture A + B 
can be met by finding a disclosure of a mixture A + B + C and a search 
for a chemical compound A B can be met by a disclosure of a chemical 
compound ABC. Since the searcher is not generally aware of how the 
subject matter is disclosed he cannot be completely certain as to where it 
is classified and so may miss pertinent answers to his questions. 

Since automatic data processing machines can sense and interpret frag¬ 
ments of a combination independently of each other, and of the combina- 



270 


PUNCHED CARDS 


tion, mechanization is expected to aid considerably in the solution of a 
search problem. If, for example, a given disclosure is analyzed into a rela¬ 
tionship A + B + C, and this information is transcribed to the machine 
according to a logic which defines the relationship, the machine can be made 
to recognize A, B, A + B, and so on, each independently of the rest of the 
context. The problem from the point of view of mechanization is to select 
these units A, B, C, etc., to serve as “building blocks” and construct a 
logic for definition of the units and their relationships. These “building 
blocks,” or “descriptors,” as they may be called, would be recorded and 
manipulated by the machine according to a logic whereby the disclosure 
could be reconstituted in a manner congruent with the logic of the search 
requests. 


Schedules of Descriptors 

Pending the accumulation of sufficient experience to make an optimum 
selection of descriptors, the initial selection is done on a tentative and 
approximate basis. For chemical compounds the descriptors were of the 
same type already found useful for searching by conventional means. The 
functions were selected as a result of the analysis of the patent disclosures 
on the basis of educated guesses as to the terms most likely to be used in 
searching. A schedule of descriptor terminology was built up which con¬ 
sisted of four sections: the inorganic and organic sections, and the “com¬ 
plex” section which contained a classification of plants, animals, minerals, 
and terms deemed pertinent for identification and searching of the mate¬ 
rials. The fourth section contained terms of function and included a classi¬ 
fication of disease in terms of body systems and pathogenic organisms. The 
terms were generally arranged in order of decreasing genericity, indicated 
by indentation. Thus, 


Heterocyclic Compounds 
Para-n-benzene Sulfoxy 
Azoles 
Thiazoles 
Oxazoles 


1313 

1313-1512 

1313-2512 

1313-2512-1423 

1313-2512-1523 


The codes reflected the same pattern of indentation. The most generic 
term of any particular aspect was called a “first position” term; the next 
indented term was a “second position” term and so on. The code of an 
indented descriptor contains the codes of all descriptors generic to it. Thus, 
“Thiazoles” contains the code for “Azoles” and for “Heterocyclic Com¬ 
pounds.” Since the generic class descriptors were inherently within the 
more specific class descriptors a search by a generic term would retrieve 
disclosures of all materials falling within that class. 



CLASSIFICATION, SEARCHING AND MECHANIZATION 


271 


Organization of Disclosure For Coding 

Each disclosure of a composition was arranged according to an organi¬ 
zation of two levels, which may be symbolized as follows: 

I • [(A B C D) (E F G) (H I)] 

where the alphabetical symbols are the descriptors. The first level asso¬ 
ciated the descriptors within parentheses and the second level associated 
the parenthetical groups within brackets; similar to the association of 
letters in a word and words in a sentence. The collection of descriptors 
within each parenthesis represented an item and the collection of items 
within each bracket represented a mixture of items or composition. 

In coding this disclosure each descriptor within the item group repre¬ 
sented a different aspect of the same material or function. If, for example, 
“sulfathiazole” was one of the items, A would represent “a thiazole,” B 
the term “a sulfonamide” and C “an aromatic amine.” If it was disclosed 
that the sulfathiazole was a bacteriostatic, D, this function descriptor was 
also associated with the other descriptors within the same level. 

The disclosure “olive oil, having a solvent function,” can be represented 
as (E F G) where E is “a fatty acid ester,” F is “an extract of the olive 
plant” and G represents a “solvent.” 

The function of the composition was represented as a separate item. 
(H I), for example, would represent the function, “tonsillitis,” H being 
‘‘a disease of the mucous membranes” and I being “a streptococcus infec¬ 
tion.” 

Thus, formula I above, exemplifies the organization for coding the fol¬ 
lowing type disclosure “a mixture of sulfathiazole and olive oil, wherein 
sulfathiazole is a bacteriostatic, olive oil is a solvent and wherein the com¬ 
position is for the treatment of tonsillitis.” 

The Punched Card 

The punched card used is illustrated in Figure 12-1. The first ten fields 
of the card were allotted for punching the document identification. Col¬ 
umns 13 to 80 inclusive were divided into seven sections corresponding to 
the seven positions of descriptors as they appeared on the schedule. A 
code represented one descriptor of the schedule and it was sectioned into 
its corresponding positions. Thus, for “thiazoles,” a category in the 3rd 
position, the 1313-2512-1423, was punched in any horizontal row (shown 
in row 7 of Figure 12-1 in sections 1, 2 and 3, respectively. In scanning by 
machine, each horizontal row was sensed independently of any other row 
and the information was recognized as such regardless of the row in which 
it appeared. The codes pertaining to an item were associated with each 
other by a punch at the upper end of the group in column 12. The scanning 



272 


PUNCHED CARDS 


PATENT 

I I 

I I 

I > I • I • I I • 4 
Ilf I III 111 


2222222222 

2 3|3 3|3 3)3 

4444444444 

9SS((S$SS9 

lllllllll 

(77M 7 7 ( 7 7| 

lllllllll 

iminti 


nmumonmmmna 

I || I I 1 ||1 I 1 v| 




I" POSH. ■ 

I II 

I I II II I II 


puiMilir 

j"l"||u 


ii||tii|iiu 

j|i|nn>||> 

JJII<<I<<II<44 
SS|liS||lt 

<Hi“i<<‘ii*» 

*I»>J|J|M 

|Mii|u|ini 

??*l*!l*!!l!»* 


C 3 "P 0 SN. 

I II 

I II 

ll|M4||ll| 

Mwammuammm 

|1 I I I 1 I I 1 1 I I 


2|2 2 2 1|22 
3 31|)||3 3 
444|||44 

M|7||77 
mlUliB 

•IS2IIV 


4414 


47|7 


Jll 


I! 




pinni 

«•»» van 
111 I I II I 


*i< 


tr 


i |i n|> i||i 

p>|n)i||np 

44444||4 

4«II<>II|I<>M 
!>|m||n[ 
lllllllll 

II!!*!!!! 


4' 


vmamumm 

[lllllllll 




[22222222 2 ] 

3 3 3 3 3 |3 
4|444 

ssss|ss|s 
11111(1 
[7 7 7 7 7 7 7 7 

lllllll 

1111119 


[lllllllll 


|22222222 
3333333 
4444(4 
I9999SS 
lllllll 
7777777 
lllllll 
Hill II 


2222222 
3333333 
444444 
SS999 
llllll 
7 7 7 7 7 7 7 
lllllll 
9999919 


IP 


ir 


ip 


44k 


III! 

[I I I I I 
2 22 2 
3 3 3 3 
4444 
S|S IS 9 S 

1111 

7777 

nn 

mi 


Figure 12-1. 


was continuous from card to card and the association of items in a compo¬ 
sition was signalled by the 2nd level punch in column 11. The last card of 
each group pertaining to a composition was selected if a hit had been regis¬ 
tered, and the identifying data on the card indicated the source of the 
disclosure. 


Searches 

This coding organization permitted the finding of the disclosures in 
terms of any one or more of the descriptors regardless of the presence or 
absence of the other descriptors. The above hypothetical composition 
might be in conventional classification under sulfa drug medicines—Class 
167, Medicines, Poisons and Cosmetics, subclass 51.5. A searcher for “an 
aromatic amine in admixture with a fatty acid ester” would find no obvious 
reason to search Class 167, subclass 51.5, yet said location would contain 
a fully pertinent reference for the desired search. A few examples of types 
of searches which would successfully retrieve the illustrated disclosures 
are: 


1 (A) A compound of the thiazole class 

2 (A C) A thiazole—aromatic amine compound 

3 ((A) (G) (H)j A thiazole plus a solvent for use in diseases of mucous 

membranes 

4 ((I)) A composition for treatment of streptococcus infection 

regardless of what the ingredients are. 


Performance Tests 

The tests were performed with respect to actual patent applications. The 
441 patents containing 6,272 items described in terms of 18,650 descriptors 
were searched in 4.5 minutes or about 95 patents per minute. The princi- 


























CLASSIFICATION, SEARCHING AND MECHANIZATION 


273 


pies described were successfully embodied. The card-sorting machine was 
a temporarily modified I.B.M., E.S.M. 101. Further details of this project 
have been previously published* • 17 . 

THE SECOND EXPERIMENT 

The first experiment indicated the feasibility of mechanizing the patent 
search and it was subsequently decided to expand to an operational basis 
for the entire chemical arts. In view of this broader program, new princi¬ 
ples had to be developed to provide greater flexibility and effectiveness 
than was available according to the earlier project. Some of these develop¬ 
ments, described as the second experiment, are being jointly undertaken 
with the National Bureau of Standards. 

The description necessarily presents various segments of the over-all 
problem. Integration into a more unified and comprehensive picture is 
expected as the developments progress closer to completion. Related 
progress in the nonchemical arts has been described 18 . 

The Descriptors 

The descriptors used in the first experiment were generally of the “com¬ 
binatory” type. For example: 

1. Amine-containing compounds 

2. With hydroxy groups 

3. Aromatic 

Category 2 does not involve more specific delineation of category 1 but 
involves instead a combination of category 1 with a group extraneous to it. 

This type of schedule has certain advantages in that it sets forth relation¬ 
ships between different chemical groups. Thus category 3, by definition, 
sets forth a benzene structure containing both an amine group and a hy¬ 
droxy group. The disadvantages, however, reside in the lack of genericity 
provided for categories 2 and 3. A search for hydroxy compounds or aro¬ 
matic compounds, regardless of the presence or absence of other chemical 
groups, cannot be made according to this particular example. 

The second system attempts to provide many more generic search as¬ 
pects, while at the same time retaining the various relationships. In the 

17 Bailey, M. F., B. E. Lanham and J. Leibowitz, “Mechanized Searching in the 
U. S. Patent Office,” J.P.O.S., 25, 566-587 (August 1953). (Copies are available from 
Research and Development, U. S. Patent Office, Washington 25, D. C.) 

'* Andrews, D. D., and S. M. Newman. “Storage and Retrieval of Contents of 
Technical Literature, Nonchemical Information,” May 15, 1956. (Copies are avail¬ 
able from Research and Development, U. S. Patent Office, Washington 25, D. C.) 



274 


PUNCHED CARDS 


present schedules any indented descriptor refers to further specificity of 
the broader descriptor rather than a combination, as 

Non-metal 

Metal 

Light metal 
Heavy metal 

The relationships among descriptors are shown by special devices which 
will be described. 

Chemical Compound Coding 

The structural formula representation of a chemical compound contains 
intrinsically the configuration of many chemical classes. Given a particular 
formula, the chemist can, on inspection, recognize any and all classes to 
which it belongs, insofar as these classes are definable in terms of an ele¬ 
ment configuration. By getting the machine to perform the same inspection 
and recognition, a compound would be available in terms of any structural 
class inherent within it without the need to preassign descriptors to it. 
The descriptors are thereby potentially available, in effect, to be synthe¬ 
sized as needed. 

The meaning portrayed by the formula can be conveyed by a descrip¬ 
tion of each element in the formula and its connectivity to other elements. 
This permits the finding of any compound in terms of any combination of 
elements in any structural arrangement within the molecule, independently 
of any other fragment or of the complete structure. Several methods for 
coding in this element by element fashion have been developed; determi¬ 
nation of the optimum method is expected after adequate machine tests. 
The methods depend on what is called the “interfix” device. 

The Interfix 

The interfixes are numerical descriptors wherein significance is in relative 
rather than absolute values. For example, in the fragment C^O-C connec¬ 
tivity is shown by sameness of interfix numbers, i.e., those elements which 
have the same interfix numbers are connected to each other. The fragment 
can be coded (Ci) (O 1 O 2 ) (C 2 ). The same connectivity is also indicated by 
(C 6 ) (Ob. 9 ) (C 9 ). 

By this method, numbers are assigned to each connection between 
elements. The numbering may start at any point in the structural formula 
and is entirely random insofar as sequence is concerned. Elements which 
have the same number are connected, regardless of the absolute numerical 
value. 

A variation of this “element by element” coding concept, which has 



CLASSIFICATION, SEARCHING AND MECHANIZATION 


27 a 


been successfully tested on the SEAC at the National Bureau of Standards 
involves the following method 19 . 

The configuration 


C 4 


—C—O—C—O—C— 

1 2 3 5 6 

can be described (disregarding the types of bonds) by saying, 

# 1 C is connected to #2 O 

#2 O is connected to # 1 C and m3 C 
#3 C is connected to #2 0, #5 O, and #4 C 
#4 C is connected to #3 C 
#5 O is connected to #3 C, #6 C 

# 6 C is connected to # 5 O 

From this information the element configuration or any part thereof 
can be reconstructed. This would hold true no matter what numbers are 
assigned to the element, just so long as each element is uniquely identified. 
For example 


C—O—C or C—O—C 
1 2 3 2 1 3 

The same group is numbered two different ways. The first will be coded 
to mean 

# 1 C is connected to # 2 O 

#2 O is connected to ml C, #3 C 
#3 C is connected to #2 0 

The second 

#1 O is connected to #2 C, #3 C 
#2 C is connected to #1 O 

# 3 C is connected to # 1 O 

Reconstruction of each of these different codes will yield the same struc¬ 
tural fragment. 

Another method for coding chemical compounds involves the use of two 
structural entities, the ring configuration and the chain configuration, to 
constitute two major building blocks. The connectivity of the elements in 
the ring as well as in the chain is shown by the coding sequence for each 
building block. Juncture among the blocks is indicated by two types of 

'* Ray, L. C., and R. A. Kirsch, “The Use of Automatic Data Processing Systems 
in the Retrieval of Technical Information,” National Bureau of Standards, a pre¬ 
liminary report. Publication is expected. 



276 


PUNCHED CARDS 


such structural interfix, i.e., the “shared element” interfix and the “bond 
juncture” interfix. The former applies to such joinings as exist in fused 
ring and spiro arrangments; the latter refers to attachments of chain to 
ring or ring to ring as in diphenyl. Additional descriptors are used to de¬ 
scribe any configurations not intrinsic in any specific element arrangement, 
such as “amide,” “acid,” etc. More detailed accounts of this and other 
fragments of the over-all problem have been published 11 . 

Organization of the Disclosure 

Three levels of organization of the disclosure for coding are used. 

II {[(AtF A ) (BiFb)] [(Ci.«Fc) (D*F d )] [(E 2 F e ) (G 2 F q )]} 

The first and second levels are groupings for the item descriptors and the 
items, respectively, as in the first project. The third level indicated by the 
bowed brackets is the process level. For convenience in exposition a single 
alphabetical symbol is used to represent all the descriptors for each item, 
except for function. The function is indicated by F and the subscript to F 
indicates what material it is a function of. 

The numbers are “sequence” interfixes. A higher number indicates a 
later step in a reaction. Thus formula II, in terms of process, may be read 
as follows: 


(1) A + B -> C 

(2) C + D —► E + G 

In addition to the process, the formula may indicate three separate com¬ 
positions, each containing two items. Thus by the interfix, relationships 
can be shown which cut across the groupings. C and D are in the same 
group as ingredients of a composition, but in different groups insofar as 
process steps involving the composition are concerned. The indications as 
to which are starting materials and which are results of the process are 
shown by descriptors associated with the ingredients. 

A process, “N1 PI Q2 R3,” symbolizes that N and P are carried on si¬ 
multaneously, and that reaction N and P each precede Q and R. 

The function never appears as a separate item but is always on the first 
level associated with the ingredient. Where the function is related to a 
particular composition it is distributed as a descriptor to each ingredient 
of the composition possessing said function. Thus, if A plus B is an insec¬ 
ticidal mixture and on the addition of C, the insecticidal property is no 
longer existent but the new admixture functions as a herbicide, this dis¬ 
closure is symbolized in formula III. 

[ (AIpH p ) (BI p Hp) (CHp) ] 


III 



CLASSIFICATION, SEARCHING AND MECHANIZATION 


277 


where I is “insecticide” and H is “herbicide” and “p” indicates a partial 
or shared function, which function is shared with any other ingredient 
having the same function descriptor. 

Alternatives 

Many disclosures of the so-called alternative type are found in patents. 
For example, “A is mixed with B or C”. Several relationships are thereby 
shown. 


(1) A + B 

(2) A + C 

(3) C is alternative to B for use in admixture with A 


Not shown is 


(1) B + C 

(2) A + B + C 

The disclosure should not be selected in a search for B + C. Another “al- 

X 

/ 

tentative” situation is found in chemical formulas of the type R where 

\ 

Y 

X is one of a, b, c and Y is one of d, e, etc. 

The searcher cannot, of course, have any foreknowledge of the existence 
of an alternative situation with respect to the particular combination in 
which he is interested. He cannot specify, as a practical possibility, “A + 
B but not if A is alternative to B” since the alternativeness may exist on 
any level with respect to any element or group of elements. The “alter¬ 
native” situation is therefore handled by a special signal, grouping the mem¬ 
bers of any alternative group, wherein an automatic discrimination is made 
for the proper selection according to the logical rules embracing the alter¬ 
native situation. 

Modulants 

The modulants are ways of getting more versatility from the schedules 
and showing many more relationships. For example, the following is a set 
of modulants. 


Disease of 50 

Disease by 60 

Infection by 60-10 

Toxicity by 60-20 

Ingredient 70 



278 


PUNCHED CARDS 


If the term for a particular plant such as a mold is associated with code 
60-10, the term has been modulated to express the idea of a disease by 
infection with that mold. If code 50 is associated with the mold, it is thereby 
indicated that the mold itself is diseased. Code 70 associated with the mold 
indicates the mold to be an ingredient in a composition. Code 60-20 asso¬ 
ciated with a particular material expresses the idea of a disease of poison¬ 
ing by that material. Thus, the modulants are used to inflect or vary the 
meaning of the root terms on the schedule. 

Negatives 

Certain disclosures are expressed in negative form such as “phenols 
may be used except those containing halogens.” Provision has been made 
for finding the disclosure of positive assertions as to the absence of chemi¬ 
cal groups. 


SEAC TEST 

A test of the structural formula search method has been made, as previ¬ 
ously indicated, using the SEAC at the Bureau of Standards, and prepara¬ 
tions are being made to test the logic of the comprehensive system involv¬ 
ing the entire chemical field. 20 - 21 The computer will be used to simulate a 
searching machine and it is expected, as a result of these tests, to evaluate 
the principles and logic of the system and to determine the requirements 
of an optimum search machine. 

10 B. E. Lanhan, J. Leibowitz, H. R. Roller and H. Pfeffer, “Organization of Chem¬ 
ical Disclosures for Mechanized Retrieval.” Presented at 131st ACS Meeting, Miami, 
Florida, April 8, 1957. (Copies are available from Research and Development, U. S. 
Patent Office, Washington 25, D. C.) 

** H. Pfeffer, H. R. Roller and E. Marden, “A First Approach to the Patent Search 
on a Digital Computer (SEAC).” Presented at the 12th National Meeting of the 
Association for Computing Machinery, Houston, Texas, June 20, 1957. (Copies are 
available from Research and Development, U. S. Patent Office, Washington 25, D. C.) 



Chapter 13 

APPLICATION OF PUNCHED CARDS 
TO LIBRARY ROUTINES 


Madeline M. Berry 
National Science Foundation, Washington, D. C. 


Introduction 

Punched cards are a convenient tool for the recording of index entries 
to scientific and technical documents, and for the finding of references to 
pertinent documents in answer to given questions. The use of punched 
cards for such operations is described in other chapters in this book. These 
tools can also be applied to other tasks, such as the many clerical and 
technical routines necessary for the functioning of a library. The use of 
punched-card systems offers the possibility of performing clerical routines 
with a minimum of time and effort, of relieving the drudgery often associ¬ 
ated with these tasks, and of freeing staff time to be devoted profitably 
to more professional work. 

This chapter presents a summary of actual and possible applications of 
punched-card systems to library routines. There is no attempt here at 
exhaustive coverage. Nor is it possible to describe the systems in sufficient 
detail to enable installation of a system in a given situation. Rather the 
chapter tries to suggest ways in which punched-card systems, manual or 
mechanical, can be applied to advantage in performing many library op¬ 
erations. The reader is referred to “Library Applications of Punched Cards: 
A Description of Mechanical Systems” by Parker 1 and “Marginal Punched 
Cards in College and Research Libraries” by McGaw 2 , as well as to other 
texts and articles noted, for more complete discussion of the subject. 

Library routines may be divided into processing functions and reference 
functions. As noted above, punched-card systems have been used success¬ 
fully in literature reference work. Other chapters of this book describe 
such applications, at least in principle. This chapter will consider only the 
processing functions in libraries. These functions may be considered to 
include those steps involved in preparing books for use—ordering and ac¬ 
quisition, binding and cataloging; and those involved in the use of books— 

1 Parker, Ralph H., Library Applications of Punched Cards: A Description of 
Mechanical Systems, Chicago, American Library Association, 1952. 

* McGaw, Howard F., Marginal Punched Cards in College and Research Libraries, 
Washington 7, D. C., Scarecrow Press, 1952. 


279 



280 


PUNCHED CARDS 


circulation. Personnel and financial administration routines in libraries can 
be performed by punched-card systems as in industrial organizations, so 
they will not be considered here. 

Ordering and Acquisition 

For the performance of ordering routines, punched cards have been used 
to replace the purchase order form typed and filed in multiple copies. The 
cards, manual or mechanically-sorted, may show any or all of these items 
of information: author of the volume ordered, its title, dealer or agent 
through whom purchases are made, and date ordered. In mechanically- 
sorted systems especially, the cards may also show estimated price, order 
number, and such information. When books are received, they are checked 
against the “on order” file and the cards are punched or notched with the 
date received and perhaps the actual price. This file then can be thought 
of as the master record of volumes purchased by the library. If the cards 
contain additional data, such as subject classification of the book, age 
group to which it is applicable, the language of the book, its literary form 
(i.e., fiction or poetry), or the country of its origin, this file may help the 
library administrative staff in the molding of acquisition policy. The Mont¬ 
clair Public Library, in Montclair, New Jersey, uses its IBM-card file to 
make such analyses for policy decisions (Figure 13-1). Thus types of non¬ 
fiction bought for adults, book purchases made for children, sources of 
purchases, languages represented, and additions to special collections have 
all been listed easily by proper sorting of the punched cards. It has been 
possible, for example, to learn the extent of use and popularity of material 
according to the date of publication 3 . 

Various forms of cards have been used successfully for ordering and 
acquisition work. An IBM card system is used in the Order Department 
of the University of Florida Libraries 4 . The Boston Public Library uses a 
dual-portion IBM card for purchasing records, and prepares purchase ana¬ 
lyses reports, including accounts of net value for various funds, reports of 
the status of the City Fund, and the like. When the cards have been used 
for such tabulations for a particular period, they are sorted by title and 
author and filed in the Purchasing Department for further reference 3 . 

* Quigley, Margery, “Business Machines in a Public Library,” American City, 60, 

101-2 (1945). 

Numerous personal communications and unpublished reports were also 
made available by Miss Quigley. 

* Duer, Margaret M., and Lewis, Clark S., “How We Use IBM,” Library Journal, 

78, 1288-9 (1953). 

6 “Purchase Analysis Procedure—Boston Public Library” (mimeographed) New 
York, International Business Machines Corp., 1934. 



APPLICATION TO LIBRARY ROUTINES 


281 



Figure 13-1. Punched card used at Montclair Public Library. 


The IBM punched-card installation at the Milwaukee Public Library® 
has proved its value in book budget accounting. The file furnishes monthly 
cumulative totals of money paid out to dealers, of orders still outstanding, 
and of the total money remaining in the book budget. This operation aids 
in spreading expenditures and the work of processing books received. It 
also allows a consistent follow-up on items not received. Easy access to 
the information on book purchases, as supplied by the Tabulating Divi¬ 
sion, guides the library staff in its purchasing and discard policies. Studies 
are made of the elapsed time between order and receipt of a book, so tech¬ 
niques of purchasing can be improved and the flow of work better organ¬ 
ized. The punched-card file can even be used to handle payments of book 
and periodical invoices. Other sets of cards are used in shelf-listing. The 
breakdown of subject matter according to major classification divisions 
(by Dewey numbers) shows where holdings are inadequate, or out-of-pro- 
portion, as an aid to buying or discarding procedures. 

Some of the advantages gained by adoption of mechanically-sorted 
punched-card systems can also be obtained by use of manually-sorted 
cards. One consideration in the use of mechanical equipment is the expense 
of its installation and operation versus the amount of work to be done. 
One merit of the manual systems for small installations lies in the fact 
that the only special equipment required is a hand punch and a sorting 
needle. The University of Illinois Library 7 uses McBee Keysort cards for 
its acquisition records. An order card (Figure 13-2) is made for each title 
to be purchased. The fund on which the book is to be purchased, the agent, 

* Baatz, Silmer H., and Maurer, Eugene H., “Machines at Work,” Library Journal, 
78, 1277-81 (1953). 

7 Brown, G. B., “Use of Punch Cards in Acquisition Work: Experience at Illi¬ 
nois,” College and Research Libraries, 10, 219-20 (July 1949). 













282 


PUNCHED CARDS 



Figure 13-2. Keysort card used for acquisition record at University of Illinois 
Library. 

and the purchase order number are recorded on the card. In addition, the 
codes for author and fund are punched in. When the book and invoice are 
received, the card is coded complete or partially complete depending on 
whether all volumes of the title have been supplied, and the card is refiled 
in the orders and receipts file until the book is cataloged. At that time, the 
year of receipt is coded in the upper left-hand corner and the card is placed 
in the dead file, from which it is discarded after a period of three years. 
The system shows advantages in speeding up the processing of invoices, 
allowing efficient follow-up of orders, and making the filing of cards a fast 
and almost mechanical task. One complete step, that of checking off the 
invoice against the copy of the purchase order, is eliminated. Instead, as 
soon as the invoice is approved, it is entered as a disbursement in the ledger 
and paid. At quarterly intervals all outstanding orders are sorted by fund. 
Encumbrances on each fund can then be corrected and overdue orders are 
claimed. 

In many libraries the staff preferred to use multiple slip order forms, 
filed in several ways. It is now possible to get manual punched cards in 
multiple slip form to provide the advantages of both approaches. 

Serials ordering procedures can also be facilitated by the application of 
punched cards. Frequent sorting of the file of cards enables catching sub¬ 
scriptions before they lapse. At Pennsylvania State University, a marginal 
punched-card system has permitted great savings in time when sorting for 





APPLICATION TO LIBRARY ROUTINES 


283 


a class of serials, such as renewals in a given subject due during a given 
period and ordered through a given agency*. 

The Order Division of the Library of Congress also uses IBM punched- 
eard methods for maintaining control of purchases of serial publications 
for the Library’s collections. Order Division personnel prepare purchase 
order forms and the Tabulating Office prepares the punched cards (Fig¬ 
ure 13-3). The cards contain the order number, title, dealer code, country 
code, price per year, number of issues per year, number of copies received, 
fund charged, and date of order. Quarterly tabulations are made by dealer, 


mv. a/is/so 

L1MAKY Of CONOUSS OHM SUP 

(MbM thh ftp I* Hw (Mtwfal nppM. H nutatM b mmlbkb, pltm «*• *• mw 
m rwarao of Hilt slip and rotvm. 


UMAOC. 

u woe. 


QUOTES MtCC 


AM. "(COM. 



|| 


| 

MCCIVBO 


3 

1 UT. MICS 

MAIO MICE 

X 

1 

VOW. NO. 

i 

3 

* u. a. oovcmnmcnt mimtino orrtcsi taas <mtu 


1 



1 

* m 

" 

- i - 

■ 

■ 

LLU 

■ 


3 

L 

1 


CONTINUATION TITLE 

1 



cost 

Q 

0 

• 

VO. 


a 

TO. 

3 

■ 

■ 

u 



mTi 

mu 

11111 

22222 

inn 

44444 

SS9SS 

Mill 

11 m 

mil 

i «in 

rmrmm«& 11 no m m m o m o i m m urnn 

1111111111111111111111111111111111111111 

2222222222222222222222222222222222222222 

3333333)333333)33331)3)3)3333333333)3333 

4444444444444444444444444444444444444444 

sssssssssssssssmsssssssssssssmssmssss 

sinus sssimimmimi si mssimismiiiii 

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 7 7 7 7 7 7 7 7 7 7 1 7 

iitmsiiiiiiiiiiiiiiiiMiiiiiiiintiiii 

miminmnninminiiiJimiiitn 

• i • t a ti u o ms m ii a «»p ij tin .in r»»n »w ibki ■ va »•« qommi 

-M1HI 

U 

a 

AMI 

1111 

3223 

3)33 

4 4 4 4 

ASM 

MIS 

1111 

IMS 

MM 

OQQOl 

Ml 

*•«« 

111 

222 

33) 

444 

SSS 

MS 

7 77 

Ml 

Ml 

apt) 

M Mid 1 

I 11 Ijl 1 

22 21J22 

3 1 3 )') 3 

1 

4 4 4 414 4 

5 M SIM 

ISSVSS 

1 

7 7 7 7*7 7 
1 

1S 1 SjM 
mm!m 

MNatnaa 

Ml 

■ Nil 

1 1 1 

222 

333 

444 

SSS 

SSS 

777 

Ml 

Ml 

MM N 

M 

DM 

1 1 

M 

WM 

M 

■ ■ 

n 

DM 

MM 

M 

m a 

■ a 

n a 

Mill! 

n"n»ni 

Mint 

2 2 2 22 2 

3)3)33 

444444 

MS9SS 

MSSSS 

7 7 7 7 7 7 

SIMM 

MIMS 

4j 

I 



Figure 13-3. Order slip (above) and punched card (below) used for maintaining 
control of purchases of serial publications at the Order Division of the Library of 
Congress. 

















































284 


PUNCHED CARDS 


order number, and country, and other special tabulations are made as 
requested 8 . 

Punched cards are also used for fiscal controls of book funds. Semi¬ 
monthly listings are made for the two major appropriations by form of 
order (Regular, Blanket, and Continuation); within each type of order a 
further breakdown is made by Recommending Officer or form of material 
sub-allotment. A yearly report is also made for pieces purchased by fund, 
country, and form of material. It serves as a summation of all serial pub¬ 
lications purchased in the year. 

Cards for serial-ordering routines usually have space for coding the 
countiy of origin, frequency of publication, language, price, renewal date, 
source, subject, type, and so forth. The Milwaukee Public Library uses 
its IBM card files to analyze serial purchases as it does book orders: who 
gets what, and the strengths or weaknesses in coverage of the various 
subject areas. Such analyses aid in determining purchasing or cancellation 
policies. 

The Library at Dow Chemical Company in Midland, Michigan, uses 
an IBM card system for renewal of serial subscriptions. One set of cards 
has titles of journals, another set has addresses of vendors through whom 
the journals are ordered. The two sets are tied together with the identi¬ 
fying numbers assigned to each serial publication. When the various de¬ 
partments of the company have checked the list of journals they subscribe 
to, the address file is rearranged by vendor to facilitate the renewal pro¬ 
cedure. The purchase order then takes the form of lists of titles prepared 
from the punched cards and sent to the appropriate vendor*. 

Punched cards are also a valuable tool in keeping records for inventory 
control, and for measuring a library’s resources. The value of any missing 
books can be calculated, or the monetary value and number of all the 
books in a given subject class can be computed. The number of duplicate 
copies of documents, the number of gift volumes, the number of volumes 
in a given language, can all be listed. 

The IBM card inventory deck at the Dow Chemical Company Library 
is used to determine shelf space requirements and rate of growth of the 
library’s collection. The file is checked every three years to determine op¬ 
timum spacing on the shelves for the next period*. 

* Keller, Alton H., “Book Records on Punched Cards,” Library Journal, 71, 

1785-6 (Dec. 15, 1946). 

Personnel of the Library of Congress gave invaluable help in preparing up- 
to-date descriptions of the use of punched cards at the Library. Their assistance 
is appreciated. 

* Taylor, F. Lowell, personal communication, June 1956. 



APPLICATION TO LIBRARY ROUTINES 


285 


Binding 

Better control of material from the time of ordering until final discarding 
can be provided by punched-card files. Thus binding schedules or bindery 
records can be maintained by means of such systems. The cards might 
contain codes for the month and year of publication of the material gath¬ 
ered for binding, the title and call number, the type of publication, color 
of binding, the size, the number of volumes to be bound as one, and such 
pertinent data. The University of Georgia uses a McBee Keysort system 
for these records 2 , as does the library at Dow Chemical Company. On the 
other hand, a mechanized system can be used for the same routine, with 
the possibility of searching each month for titles scheduled to be gathered 
together and sent to the bindery. It is possible to include the information 
about binding with the information about subscriptions on a single card 1 . 

Cataloging 

The process of cataloging may also be facilitated by use of punched cards. 
This is especially true in the matter of producing catalog lists by means of 
machine-sorted cards. For example, the King County Public Library in 
Seattle assembles and prints its catalog by the IBM system 10 . The library 
has many branches, with constantly changing catalogs. The mechanized 
system relieves the drudgery and time-consuming task of removing cards 
for books sent back to the main library and of adding cards for books which 
have been newly acquired. Instead, lists of names of books are tabulated 
with the IBM equipment at the main library, and assembled in looseleaf 
style. The lists are arranged as a catalog of adult books, one of juvenile 
books, and a combined alphabetic list, plus an author catalog. The lists 
are changed about every six weeks. The master cards from which the var¬ 
ious lists are prepared include a classification number and the symbol used 
with it (e.g., B = bibliography, J = children’s book), author, title, lan¬ 
guage, reading level, and a code for the subject matter. A duplicate file, 
called the locator file, is maintained. When books are sent out to the 
branches, the card is stamped with the branch name and the date it was 
sent out. 

The simplified subject codes, punched in the master cards as mentioned 
above, are arranged alphabetically and assigned consecutive numbers. 
Spaces are left between the code numbers to allow for additional subjects 
to be included in the alphabetical list. Since these codes are general, a key 
or index is prepared to direct the user to the heading under which to look 
for a given subject (e.g., accounting, household—under home economics). 

10 Alvord, Dorothy, “King County Public Library Does It With IBM,” Pacific 
Northwest Library Association Quart., April 1952. 



286 


PUNCHED CARDS 


There is much sorting required to prepare these lists, but even so it is 
less work than required for typing of cards, filing, and so forth. One ad¬ 
vantage is that “the concentration of work at headquarters .. .(is).. . not 
a burden on the local librarians in the community branches.” The branches 
maintain collections of only 8,000 to 9,000 books, but they are ever-chang¬ 
ing collections. Maintenance of the catalog lists by mechanical means per¬ 
mits an easy and economical method by which the staff can keep up-to- 
date with the changes. 

The circulation record cards at the Milwaukee Public Library are used 
in like manner to prepare lists used as catalogs on the Bookmobile®. 

The Library of Congress prepares a continuing supplement to the Union 
List of Serials as a monthly listing called New Serial Titles , which is avail¬ 
able on a subscription basis. Punched-card methods are used in the prepa¬ 
ration of these lists. Serials first published after December 31, 1949, and 
received by the Library of Congress and nearly 300 cooperating libraries 
are arranged alphabetically. The listing (Figure 13-4) shows title of the 

660 

chemical industry and engineering. Sydney. 

1. MY 1953- 

V. 1* NO. 3-5. JL-S 1953 OUT OF PRINT* 

MONTHLY. E.G. HOLT PUBLISHERS* 166 PHILIP 
STREET* SIDNEY* AUSTRALIA. $3.00 
K U 4- N N 1- 



Figure 13-4. Listing (above) and punched cards (below) used in preparing the 
Union List of Serials at the Library of Congress. 







































APPLICATION TO LIBRARY ROUTINES 


287 


serial, place of publication, frequency of issue, the holding libraries (by 
National Union Catalog symbols) and the issue with which the library’s 
holdings began. The entries are coded on the punched cards by subject 
content, language, and country of origin, so it is possible to make general 
listings by subject, country or language arrangements, or special listings 
of serials on given subjects, from given countries, or in given languages. 
Lists in classed subject arrangement, for example, appear in twelve monthly 
issues, sold on a subscription basis like the alphabetic list. 

The alphabetic lists appear monthly and in annual cumulation which 
are self-cumulative over five-year periods. Using punched cards for the 
entries makes such annual and five-year cumulations easier to prepare and 
to print. 

Another interesting project is one now going forward in Italy, in which 
a national Union Catalog of Italian libraries is being prepared by means 
of Remington Rand punched cards. Thirteen important libraries are re¬ 
cording their holdings on cards or tapes, which are then collected at the 
National Central Library Vittorio Emanuel. The tapes are converted to 
cards, and then the cards are alphabetized, checked for duplicates, and a 
single master file produced. From this deck will be produced sets of cards 
or sheets of paper printed with the bibliographic information. These cards 
or lists will be sent to all the most important Italian libraries, which will 
then check their holdings against the holding of the Roman libraries. Adding 
to and revising the first lists will eventually result in a Union Catalog for 
all Italian libraries. 

Additional ways in which to use the information recorded on such cards 
in preparation of catalogs cannot always be predicted in advance. Such 
information might permit compilation of statistics or the conducting of 
studies which may be necessary for justification of budgets, for definition 
of policy, or for public relations work. 

Circulation 

Keeping track of the actual use of books by maintaining circulation 
records is the work area with the greatest potential for the application of 
punched-card systems. Both manual and mechanically-sorted systems are 
much in evidence in public libraries, those of colleges and universities, and 
special libraries. The use of punched cards helps reduce the clerical work 
necessary in arranging files of cards, stamping or writing identification 
numbers and dates, or sending overdue notices. Basically, circulation en¬ 
tails transcribing the borrower’s identification to a book card, recording 
the date borrowed or the date due, and filing the book card by date for 
retrieval when the book is returned. Ordinarily this is the only kind of 
record kept by the public libraries. College and university libraries, because 
of their reference function and the consequent need for more flexible con- 



288 


PUNCHED CARDS 


trols, usually maintain class records and borrowers’ records in addition to 
the date files. Punched cards make it possible to combine more than one 
such file into a single master file. This gives only one place to search for a 
charge. Some libraries keep two files—one for active books and one for 
inactive books. This still means that there are only two places in which 
to make a complete search, and usually it is necessary to consult only one 
of the files. 

Charging systems for circulation control can involve three types of rec¬ 
ords: transaction cards, call cards, and book cards. Any of these may be 
in the form of punched cards, with consequent advantages and limitations. 

A charging system that uses punched-card transaction cards necessitates 
the following procedure. A call card or slip is filled out by the borrower 
and stamped with the transaction number and date it is borrowed. A pre¬ 
punched, pre-dated and numbered transaction card corresponding to the 
call card is slipped into the book pocket. When the book is returned, the 
transaction card is removed and the book is ready for circulation. The 
transaction cards can be sorted by month and day, and those representing 
books not yet due can be filed by date. On the due date, all cards received 
as books are returned can be sorted by number and matched against a 
master deck. Missing numbers represent overdue books. Call slips with 
the same numbers can be pulled for the borrowers’ names, and overdue 
notices sent out. The advantages of such a system are that it is speedy, 
accurate and efficient. One card provides information about the book and 
the date due and only one card is used for the charging procedure. There 
is less manual labor involved and therefore fewer library assistants are 
needed for this particular job. On the other hand, the location of a specific 
book is difficult to trace. The borrower has no record of what he has charged 
out, and there is no recorded proof of a book having been returned. An¬ 
other disadvantage is that it is difficult to take inventory of the library’s 
holdings at any given time. 

The Detroit Public Library maintains such a charging system on IBM 
cards 11 . Loan slips are stamped numerically with serial loan numbers, the 
date due, agency (or branch library) name and its identification number. 
The borrower merely signs his name and address and the call number of 
the book he wishes to take. A punched card, with the same information 
punched into it as is stamped on the loan slip, is inserted into the book 
pocket as the date card. When the book is returned, this card is removed 
and the book is again ready for circulation. Such a system eliminates any 
delay in finding the book card and re-inserting it before returning the book 
to circulation. 

11 Monkevich, Edward, “Public Library Mechanizes Book Loans,” The Punched 
Card, 1, 140-2 (1952-53). 



APPLICATION TO LIBRARY ROUTINES 


289 


Punched card sets are prepared in advance for each agency (branch 
library) with the year, due date, branch identification number, transaction 
number and deck (set) number. The cards from returned books are sent 
into the tabulating room daily. The “overdue” cards are sorted out, and 
the rest of the deck set aside until the due date is passed. The cards are 
then sorted, damaged cards are replaced, the deck is compared with a 
master deck and missing cards added, a new due date is punched in and 
the set is then returned to the proper branch. The missing cards, of course, 
represent books which are overdue, and overdue notices are typed and 
mailed to the borrowers. 

The system assigns a fixed day each week as the date due for each agency 
or branch. The books are circulated for four weeks, with this fixed due date. 
This practice simplified the charge file and reduced the number of overdue 
routines. The charging system in general is a speedier, more accurate, and 
more businesslike method for controlling circulation. 

The Free Library of Philadelphia has installed the same type of charging 
system using transaction cards, but with two important modifications. The 
charge-out step is done by a photographic method. Transaction card, book 
card, and borrower’s card are placed together in a Diebold Flofilmer camera 
and microfilmed. Thus at the end of each day, the library has a film record 
of its transactions, with complete details about the book borrowed and the 
identification of the borrower. The allowed loan period is the same for all 
books. When a due date has passed, the returned transaction cards are 
checked for missing numbers which represent overdue books. The micro¬ 
film is reviewed for the names of the delinquent borrowers as well as the 
names of the books. The overdue operation for the entire Philadelphia 
library system is located at the central library. Three clerks can handle 
overdues for its 39 branches. The use of the microfilm charging method 
makes it possible to eliminate stamping and writing on book cards and 
borrowers’ cards, thus increasing the speed and accuracy of the operation. 

The other modification employed at Philadelphia is the use of small 
40-column punched cards as transaction cards (Figure 13-5). These cards, 
made by Underwood Corporation Samas Punched Card Division, measure 
approximately 2 x 4% inches. Each card bears two due dates so it can be 
reused at six-month intervals. The new procedures using punched trans¬ 
action cards and film charging are said to have released over twenty-five 
trained librarians from tedious clerical work for professional librarian du¬ 
ties' 2 . 

The Brooklyn College Library also uses a punched transaction card sys- 

“ “The Philadelphia Story” and “The Free Library of Philadelphia Has More 
Efficient Book Control With Punched Cards,” brochures published by Diebold 
Inc. and Underwood Corp. respectively. 



290 


PUNCHED CARDS 



aeASt k£p> th: owd cook pocke 

A CHARGE * DATE y 

WILL BE MADE p 

IF THIS CARO PLE 
IS LOST ; | ^ [ 

FREE 11Bn^lO' OF 
•,,PHIL\DKLNHA 


l*, 

*I|4» rv) 4o 

J! * 


iii.riatk’i him 

AHVawl 33 


1 S 0 T SI %□"» 

oavosiHi ij 1 .ggQ f\£ 
30VW 39 niflM • __ h c 

3DMVHD V i -S, LvaQr 
llHXtf >006 *1 wi L: jjjy ttlM 


KEASC KEEP TViS CARD M BOOK POCKE . 

A CHARGE DATE E 

WILL BE MADE 

IF THIS CARO DUE * j 


FREE LIBRARY OF 
PHILADELPHIA 

°i 


CP *;<•> to # o 

J • 


viHcnaaniHJ 

30 AUVUOri 33 UJ 

•ISOT SI SE 

08V0SIHUI grjQ = 

30VW 38 TTIM | 

30BVH0 V 3xva[ 

12 ttM >008 M (MO SHL d 3 > B» 3 \( 


Figure 13-5. Transaction cards used at the Free Library of Philadelphia. 


tern, with IBM cards. Sets of GOO consecutively-numbered cards are used, 
one set for each due date, with that due date stamped on and punched 
in. The operations of sorting the cards from returned books, collating them 
against a master deck to find missing cards for overdue books, and sending 
out overdue notices are the same as described above 13 . 

Call cards can also be punched cards, of either the manual or machine- 
sorted type. In a charging procedure, the call card is made out by the 
borrower, and the date due is stamped on the call card and on a slip in 
the book pocket. The call cards are sorted according to the date due, and 
are punched or notched for that date. These cards can then be filed man¬ 
ually by classification number. When a book is returned the call card is 
removed from the file. Periodically the rest of the file is sorted by due date 
and overdue notices are sent out. With manual cards of the McBee Key- 
sort variety the notches for due date can be covered up and a new date 
can be punched, corresponding to a week later, to call attention to books 
still overdue, requiring second notices. The advantages of such a system 

13 “Recruiting Library Personnel. Automation in the Library,” 41st Conference of 
Eastern College Libraries, Columbia University, November 26, 1955. Published 
as ACRL Monograph 17, Chicago, Association of College and Reference Li¬ 
braries, 1956, p. 34. 


/f£B 06 i .49.1,0284 i ,AUG-07 


APPLICATION TO LIBRARY ROUTINES 


291 


' 


FILL OUT COMPLETELY; PRESENT AT CIRCULATION DESK 


1 

IF BOOK It MOT AVAILABLE 

ntr 

CALL MM OCA 

AtfTMO* 






[ ] WILL K acuua if MOUCSTU 

[] AT iMtcirr 





TITLl 


1 

VOUMC 



Ml 

□ — 




M 

» 


corv 



i © 

s 

[3 

[] mun 



m 

1 

St 

[] MT LOCATES; mOTMCS •!*«■ 

[] CAAMATC STVMST 

T«L m ”“j 


Vltl M SAM. If WHA 

VM VUL K MTiriU 

£3 —-snout 

iisstfssi 

o 

UNIVERSITY OF MISSOURI LIBRARY 

It St 

LOCAL AOBSSSS 



is* 



J 


Figure 13-6. Punched card used as call slip at University of Missouri. 


are its speed, accuracy and efficiency. One card gives a record of the book 
and the date due. Overdues can be handled easily by simple sorting pro¬ 
cedures. Conversion from other systems is not difficult and the records are 
simple and flexible. One disadvantage of the procedure is that punched-card 
call cards are more expensive than the usual call slips. 

The University of Wisconsin maintains its circulation records by using 
IBM cards as call slips. These cards are filled out by the borrowers, and 
when the book is taken the cards are kept at the circulation desk until the 
end of the day. The date due is then gang-punched through all the cards. 
They are then interfiled manually. Twice a week, the cards are sorted and 
those which represent books overdue drop out. Notices are then sent out 
manually 14 . 

The University of Missouri also uses IBM cards as call slips (Figure 13-6). 
The original slip is placed in the card pocket of the book and is used to 
discharge the circulation record when the book is returned. Overdue notices 
are sent out weekly by making a Thermofax copy of the IBM card in the 
file. 

Book cards in the form of punched cards are used in the same manner as 
conventional book cards. When the card is a manual punched card such 
as a McBee card, the date due and other information can be notched into 
it. The book card is signed by the borrower and a date due card is slipped 
into the pocket of the book. The book cards are sorted and filed by classi¬ 
fication number. If not already so recorded the due date is punched or 
notched in. When a book is returned, the book card is put back in the 
pocket. Remaining cards can then be sorted by due date in order to send 
out overdue notices. The cards can be repunched for later notices, as de¬ 
scribed above. Such a charging system has all the advantages of the sys- 


M Ibid., pp. 33-44. 






292 


PUNCHED CARDS 


terns described above. One disadvantage of these systems, when manual 
punched cards are used, is the necessity of plugging up notches before 
recording new dates such as required for additional overdue notices. How¬ 
ever, all these systems use punched cards to reduce to a single file the 
number of circulation records usually maintained. 

The Montclair Public Library has one of the country’s most highly pub¬ 
licized mechanized circulation control systems 3 . It has been described in 
various publications, so it will merely be summarized here. Each IBM card 
represents one book or one borrower (Figure 13-7). Both types of cards, 
borrower’s identification card and book card, are made from appropriate 
portions of master cards (Figure 13-1). These master cards have the de¬ 
sired information both written in and punched in. The borrower’s card 
carries the person’s identification number, address, voting district, age, 
sex, education, and occupation. The book card on the other hand has the 




Figure 13.7. Punched cards used as book card (above) borrower’s identification 
card (below) at Montclair Public Library. 
























APPLICATION TO LIBRARY ROUTINES 


293 


■1 

in 

III 

in 

■ 


q: 

33: 

3a: 

s 

3 

CJBG3 

303300 


imK 

m i 

III 

1 1 







! «W»N1 «6l<t UT C.<? I M*M Utift —f mil 



| • I I • » • >!• • « it u ij nKiWdIu '» at Jninuda a r ji a a n o u * ■ a » ■ a «!•' o «j m « <4** «.« a u u 



Figure 13-8. Loan cards used at Montclair Public Library. Above: Record of book 
charged to borrower; below: record of book returned by borrower. 

book classification number, accession number, shelf location, branch loca¬ 
tion, general type and style, language, date of publication, price and source 
of purchase. At the circulation desk is a control machine by means of which 
the information on both borrower and book cards is assembled for repro¬ 
duction at a remote point in the loan card (Figure 13-8). This corresponds 
to a charge card and is a permanent record of the loan of a book to a bor¬ 
rower. When the loan card has been produced the borrower’s card is re¬ 
turned to him and the book card is put back into the pocket in the book. 

The return of a book is recorded in a similar manner except that only 
the book card is inserted in the control machine to produce a return card 
(Figure 13-8). Return cards are matched against loan cards and the latter 
are punched with the return date and filed as a permanent record of the 
transaction. Loan cards for overdues are pulled and notices prepared for 



































294 


PUNCHED CARDS 


mailing to the borrowers. Renewals can be made by phone, in which case 
the book accession number is used to pull the loan card from the deck 
which is sorted and filed by these numbers. Then new loan cards are made 
with the new due date recorded. Reserves can also be handled easily with 
this punch-card system. A card is separated into two parts, one of which is 
filed at the loan desk for manual checking as books are returned. The other 
part of the card is used to make a duplicate loan card which can be col¬ 
lated with the deck of return cards to locate any returned books which 
have been reserved. 

Fines and fees are also recorded by the control machine. When the trans¬ 
action cards are sorted daily, those on which fines or fees have been paid 
are thrown out and a tabulation is made of the money collected for the day. 

Uncataloged material also may be handled by the punched card system. 
The card which corresponds to the book card is pre-numbered and pre¬ 
punched with the number. The borrower writes down the title of the ma¬ 
terial and then one part of the card is used as a temporary book card and 
handled as described previously for the preparation of a loan card. 

The system has been well received at Montclair Public Library by staff 
members as well as by the public. It shows advantages of speed, since the 
recording of loans and returns is accomplished by simply pressing a lever 
at the control machine; of accuracy, since numbers are automatically veri¬ 
fied and charging errors eliminated; and of rapid turnover of books, since 
the book card is in the pocket at all times and the book is available immedi¬ 
ately upon being returned. In addition, circulation statistics are accurately 
recorded and accumulated, and stored for further analysis and study. 

Some libraries with special circulation problems have found punched 
cards to be a useful tool in maintaining control of material. For example, 
the Division for the Blind at the Library of Congress keeps track, by 
means of an IBM card system, of its 30,000 to 40,000 talking book ma¬ 
chines. Approximately 6,000 to 8,000 new machines are purchased each 
year to replace worn or out-of-date models. Fifty-four distributing agencies 
for the machines send in records of loans, transfer of custody, repairs, or 
other status of the machines to the Library (Figure 13-9). The punched- 
card system makes possible the mechanical listing of all machines charged 
to each agency, as an annual inventory record. Each agency checks this 
list against its own records and sends any corrections, additions, or other 
notations to the Library, where the punched cards are changed accord¬ 
ingly. Listings can also be made for salvaged and discarded machines, 
arranged by model and serial number. Analysis of the listings reveals, for 
example, areas with excessive storage of machines, indicating that reader 
demand is not heavy enough to justify additional machines or that some 
machines registered there can be transferred to busier areas. 



APPLICATION TO LIBRARY ROUTINES 


295 



Figure 13-9. Punched cards for maintaining control of talking book machines at 
Library of Congress. 


The Loan Division of the Library of Congress has another unusual use 
for punched cards. They are used to prepare lists for recall of material 
borrowed by the libraries of various Government agencies in Washington. 
The cards contain L. C. classification number, author, title, and place of 
imprinting of the material borrowed, plus the borrower’s code number and 
the date the material was borrowed (Figure 13-10). These cards are pre- 





















296 


PUNCHED CARDS 


F7 

THE AlIjAE VI ErtMB 


■ 




1 

ASSiriCATlOM NUMBER 

1 

« 

! 

1 

i 

DATC 

j 

1 

1 

1 



* 0 A 

II II ^ 


‘“'•1 V 1 

1 1 1 1 1 1 


1 

II III 1 


1 



—f*‘« I | 

. . mi 

v ■ a ■ ii if|u m aL w|« a|» n|n ?J* j» a n * 4b 


Figure 13-10. Punched card used for recall of material borrowed from the Library 
of Congress. 


pared at the Loan Division office for each book before it goes out on loan. 
Four copies are filed manually: one as a shelf-list; one by the borrower’s 
code, sub-filed chronologically; one chronologically; and the fourth in the 
Library’s central charge file. Before being sent to the central file the latter 
are cut down in size to eliminate the borrower’s code punching. This re¬ 
tains as confidential the information about who borrows what. The file 
arranged chronologically has proved valuable as a tool for analysis of ac¬ 
quisition and discard policies. The Loan Division states, however, that 
the IBM card system was developed and is maintained primarily because 
of the ease of preparing recall lists. Since about 100,000 loans are made per 
year, sending out overdue notices is a tremendous task. Every three weeks 
the cards remaining in the borrowers’ file in the subsection for the current 
return date are pulled and tabulated in separate lists, one for each bor¬ 
rower. These lists then are merely slipped into envelopes and mailed out. 

Many libraries, both large and small, use hand-sorted punched-card sys¬ 
tems for circulation work. As stated above, punched cards obviate the 
necessity of maintaining more than one file. For example, two files, one of 
book cards arranged by call number, and one of call slips (made out by 
borrowers) arranged by due date, can be combined in one file arranged by 
call number but notched by due date for easy sorting. Thus the cards can 
be easily pulled when books are returned, and needled for books which are 
overdue. Some systems use small edge-punched cards with holes for due 
dates, such as three days for each of a number of successive weeks (three, 
four or five weeks). Direct punching records the date due. The cards are 
sorted on three days, the cards drop out for books not yet returned, and 
overdue notices are sent out. The cards are punched for the same day in 
the next week, so that they will drop out then if the books have still not 











APPLICATION TO LIBRARY ROUTINES 


297 



Figure 13-11. Call cards used for control purposes at University of North Carolina 
Library. 


been returned. The needling, sorting and refiling of cards are accomplished 
in short order. One example 2 shows 50,000 cards sorted by a staff assistant 
in hours. Arranging such a file by first, second, and third notices takes 
fifteen minutes; rearranging and refiling by call number requires one-half 
to one hour. 

The University of North Carolina Library uses colored clips to indicate 
cards which have been sorted as overdue and for which notices have been 
sent. There is then no need to sort for the relatively few overdues requiring 
second notices. The cards used in this system are not perforated at the top 
(Figure 13-11), so the clips don’t get in the way of the holes or interfere 
with sorting operations. A smooth, flat metal clip must be used to prevent 
cards from catching on other cards 2 • u . The clips are also used on cards for 

“ Hood, M., and Lyle, G. R., “New System of Book Charging for College Libra¬ 
ries,” Library Journal, 65, 18-20 (Jan. 1940). 



298 


PUNCHED CARDS 



Figure 13-12. Card used for circulation control at Wayne County Public Library. 

books to be held for other borrowers and for books currently on the “hold” 
shelf. 

The Mill Valley Public Library in California uses McBee Keysort trans¬ 
action cards and pre-dated charge slips for its circulation operation 1 *. The 
system is less costly than a mechanized operation but speedier than con¬ 
ventional procedures. Clerical and professional duties are separated, and 
the library staff is free to do more professional work. The system has the 
disadvantages of charging systems in general as discussed above. 

The Wayne County Public Library in Detroit also uses a manual 
punched-card system for circulation control. 1 * Eight serially-numbered 
decks of McBee cards are used for loan or transaction cards; each deck has 
a different colored edge, one color for each week (Figure 13-12). Charge 
slips made out by the borrower have space for information about the book 
and the borrower. A pre-dated (by colored edge) and pre-numbered (edge- 

18 Geer, Helen T., Charging Systems, Chicago, American Library Association, 1955. 



APPLICATION TO LIBRARY ROUTINES 


299 


notched) loan card is placed in the book pocket and the card number copied 
on the charge slip. These are filed in numerical order. When the book is 
returned the color on the loan card shows if the book is overdue. Returned 
loan cards are sorted by color and then each colored deck is sorted numer¬ 
ically. The file is then checked against the file of charge slips. Missing 
numbers represent overdue books, and notices are sent out. This system 
requires all books to be due on a given day in the week. The advantages 
and disadvantages of the system are those described previously for charg¬ 
ing systems. 

The Library at Wisconsin State College in LaCrosse installed a Keysort 
charging system on a two-year trial basis. The staff has found that the 
new circulation procedure takes less than half the time used with the old 
system, and that errors have been greatly reduced. One disadvantage of 
the system, it is reported, is a “certain cumbersomeness when a student’s 
library record, when withdrawing from school, has to be cleared”. It is 
necessary to check through the entire student part of the classification file. 
Even so, the average time for checking a withdrawal is ten minutes 17 . 

Cards used for such systems as described above are small, ranging from 
3 x 5 to 334 x 6 inches. There is space for the borrower to write in informa¬ 
tion about the book being taken. Holes, one or two rows, appear on two, 
three, or all four sides. 

Renewals are handled by re-dating the original call cards. The first due 
date is covered with a correction sticker and the new date is punched in. 

Punched-card files can include other than call number and date due files. 
Holes can be assigned for a faculty file, reserve book file, bindery file, etc. 
Some libraries do not keep inactive charges on punched cards, to avoid 
adding infrequently used cards to the file. Some libraries divide the punched- 
card file into active and inactive charges, or faculty and student loan, as 
use or convenience dictate. Other libraries use different colored cards in¬ 
stead of separate files. Again, some libraries use punched-card systems only 
for books going out for home use. Books to be used in reading rooms or 
study carrels have paper call slips or conventional call cards. Many special 
forms of cards and special uses of manual punched-card charging systems 
are discussed by McGaw 2 and by Stokes 18 . 

Analyses and Studies 

The application of punched-card systems to library routines results in 
faster and more accurate processing, as discussed above. By making clerical 

17 Hocker, Margaret L., “Punched-Card Charging System for a Small College 

Library,” College and Research Libraries, 18,119-22, 131 (March 1957). 

18 Stokes, Katherine M., “A Librarian Looks at Keysort,” Library Journal, 72 

(June 1947). 



300 


PUNCHED CARDS 


tasks easier, punched-card systems enable staff members to do more pro¬ 
fessional work. One of the benefits to be derived from this additional time 
is the ability to analyze the library’s operations and to study use of the 
library’s holdings. Such analyses and studies should yield revealing statis¬ 
tics to help the library improve its services. 

The Montclair Public Library has used the IBM system extensively for 
determining the types of books which are being used, where most of the 
borrowers live, which occupational groups are being reached adequately, 
and such data. These facts help the library to plan book purchases, to 
consider location of branches, and to study the reading needs of its bor¬ 
rowers. The Milwaukee Public Library staff uses its IBM set-up to corre¬ 
late book circulation with information about borrowers, to determine what 
effect age, sex or education have upon reading taste. A shift of emphasis 
in purchasing policy can then be effected if necessary. Analysis of total 
circulation by major classification divisions of the Dewey Decimal System 
also helps in formulating purchasing policy. Discard analysis is aided by 
listings arranged by Dewey classes, showing where discarding is heaviest. 
If the listings are further divided into groups according to the age of the 
materials, it is possible to determine total holdings in each age group. 

Bibliographies of the library’s holdings can be prepared by sorting 
punched cards by subject classification. Materials in a category that cuts 
across conventional department lines are then compiled for a complete list. 

Another use for punched cards has been proposed, that of helping to 
prepare a cost of books index. The work would involve developing a stand¬ 
ard of measurement for book prices for various key years according to 
subject groupings in terms of a selected base period expressed by index 
numbers. It is proposed to use 1947-1949 book prices as a base period. Such 
a price index could be used by libraries for planning acquisition policies, 
for budget justifications, and the like. A Committee on Cost of Library 
Materials Index of the American Library Association is working out the 
details of the proposed project. A tentative punched-card design has been 
suggested (Figure 13-13). The left side would be checked by the person 
examining a bibliography for eligible items, and the right side would be 
used for punching the data. One card would be prepared for each eligible 
book. The use of punched cards would facilitate the central tabulation of 
data developed at different locations. 

Punched cards have also been proposed as a means of facilitating prepa¬ 
ration and management of a Union Catalog of Serials to be established at 
the Library of Congress. The problems of defining the limits of such a cata¬ 
log and obtaining cooperative effort in preparing it have been studied by 
the Joint Committee on the Union List of Serials. 

The proposed catalog will contain a listing of the titles and volumes of 



APPLICATION TO LIBRARY ROUTINES 


301 







issue l 

item | 

year 



country 


reference 

ph, '£Y?Efo & 

n 

natural 

sciences 

u 


1 


m 

— 

m 

number pages j 

fine arts 

m 

warn 

13 


1 


m 

mm 

14 

cost 

social 

sciences 

& 

— 

medicine 

mm 








bd. 

I unbd. 



0 


16 

1 binding_I 

literature 

T 

agriculture 

or 




m 


m 

question 

novel a 

m 


19 

r 



m 


mm 


1 




77777727 Z 

8 . 

3*3 3 3 3 3 3 3 1 5 - 
§ 

4 4 4 4 4 4 4 4 4 jr 

[5*5 5 5 5 5 S 5 S “ 



99999999 

»! M * It ••»!»* 


Figure 13-13. Card proposed for use in preparing a cost of books index by a Com¬ 
mittee of the American Library Association. 


serials held by the principal research libraries in the United States and 
Canada (about 500,000 titles and over 50,000,000 volumes). As a by-prod¬ 
uct of the catalog a Union List of Serials would be published every twenty- 
five years and in-between current service with five-year cumulations. In 
addition special subject fields can be compiled. A general list by country 
of origin would also be useful, in light of the growth of area studies. 

Punched cards would be used during the period of collection, collation, 
and listing of the great majority of titles. Lists would then be sent out to 
cooperating libraries for addition of their holdings. The information se¬ 
cured in this fashion would be consolidated and typed in the center por¬ 
tion of IBM cards, leaving the end sections of the cards free for coding 
subject and country codes, and numeric codes for maintaining the titles 
in alphabetical sequence. It has been suggested that an IBM electrostatic 
printer might be used to reproduce duplicate decks of cards, if needed, or 
to run off a limited number of special lists, and to provide copy for use as 
the basis for photo-offset printing of the main list. Since the National Union 
Catalog symbols for holding libraries would be typed in the center portion 
of the cards and not in the end sections, information about the holdings of 
a group of libraries in a particular region of the United States would be 
difficult to obtain. Preparation of regional lists would probably require 
the use of additional files of punched cards with the symbols for holding 
libraries coded into them. To conserve space these subsidiary records could 
be stored on magnetic tape. 

As punched-card systems are used in more and more libraries, additional 
uses will be found for the data that can be thus handled so conveniently. 
For example, numerous libraries in organizations or institutions which al¬ 
ready have a battery of punched-card machines are using these machines, 
















302 


PUNCHED CARDS 


rent free and without a budget item for the operators’ time, for sorts and 
summaries in one phase or another of their routines. The application of 
general business methods will help a library to render its services more 
effectively. 


General References 

Kilgour, F. G., “New Punched Cards for Circulation Records,” Library Journal , 
64, 131 (Feb. 15, 1939). 

Maxfield, D. K., “Library Punched Card Procedures: Past Experience and Future 
Possibilities,” Library Journal , 71, 902-5, 911 (June 15, 1946). 

Baehne, G. W., ed., Practical Applications of the Punched Card Method in Colleges 
and Universities , New York, Columbia University Press, 1935. 

Parker, Ralph H., “Adaptation of Machines to Book Charging,” Library Trends , 
pp. 35-41 (July 1957). 

Circulation Control for Libraries } New York, International Business Machines, 
1953. 

Geer, Helen T. “Charging Machines,” Library Trends , pp. 244-55 (Oct. 1956). 

Library Literature f New York, H. W. Wilson Co. Published quarterly; bound 
annual volumes and three-year cumulated volumes also published. Contain author 
and subject index to literature on library science and librarianship. See headings 
“Machines and the Library,” “Punched Cards—Library Uses,” and “Equipment 
and Supplies.” 

Lowe, Ruth K., “Additional Bibliographic Uses for Keysort Punch Cards,” 
Library Journal , 76, 196 (Feb. 1, 1951). 

Gull, C. D., “Summary of Applications of Punched Cards as They Affect Special 
Libraries,” Special Libraries y 38, 208 (Sept. 1947). 

Smith, S. W., “- And a Few Machines,” Library Journal , 74, 1044 (July 

1949). 

Rift, Leo R., “An Inexpensive Transaction Number Charging System with Book 
Records,” College and Research Libraries , 18, 112-18 (March 1957). 



Chapter 14 

REVIEW OF APPLICATIONS 


Barbara L. Haksteen 

National Science Foundation 
Washington, D. C. 


Introduction 

Research in a specific field invariably involves the need for information 
which is related to that previously found only after a systematic combing 
of abstract periodicals and other secondary sources. In order to avoid a 
repetition of this effort and to obtain information quickly, many groups 
and individuals are turning to punched-card systems. It is the purpose of 
this chapter to present brief summaries of published articles and reports 
which illustrate the way in which a surprisingly wide variety of fields have 
defined and solved their information problems by adapting the basic prin¬ 
ciples of punched cards to suit their needs. More detailed information may 
be found by consulting the references cited. 

Nuclear Data 

IBM equipment has been used to establish an index referring physicists 
from nuclear properties to the nuclides possessing them. An IBM card is 
made for each property (half-life, energy of the various emissions, stability, 
kind of emissions from active isotopes, availability from AEC, presence of 
natural radioactivity, etc.) and each nuclide occupies the same punching 
position on each card. Two sets of different colored cards can be used, one 
to indicate light nuclides and one to indicate heavy nuclides. The system 
is indefinitely expansible since any number of properties can be included 
simply by starting a new card. Selection consists of pulling cards for given 
properties; the matching holes represent desired nuclides. The entire system 
is inexpensive since the cards are manually operated 1 . 

A Keysort card system has also been proposed which reverses the pro¬ 
cedure and has one card for each nuclide on which are punched all the 
properties it possesses. This has the advantage of having all the available 
information on a particular nuclide on one card. As with the IBM equip- 

1 Wachtel, Irma S., “A Punched Card Index for Nuclear Data,” Am. Doc., 3, (1), 
56-7, (Jan. 1952). 


303 



304 


PUNCHED CARDS 


ment, any type of information can be indexed—general statements as well 
as specific measurements 2 . 

A new research tool has been developed by Bonino and Laing, Pittsburgh 
Glass Co. Research Laboratory, Creighton, Pennsylvania. It is called the 
Raychronix Punched Card Nuclide Identifier, and uses 5x8 inch McBee 
Keysort cards. This system enables fast nuclide selection or identification 
by atomic number, mass number, chemical symbol, stability or radio¬ 
activity, availability, types of radiation, half-life and energy of radiations. 
One set of holes on the card carries the alphabet so that the symbols of the 
elements can be coded. Both symbols for elements with two names, such 
as Ra 224 and ThX, are used. The atomic number (Z) and the mass number 
(A) are coded in a series of fields so that the card for a nuclide with a known 
A or Z can be searched for and selected from the randomly ordered pack 
of cards. An example of the particular value of this tool is in the field of 
health physics where it is desirable to identify unknown radioactive mate¬ 
rials so that necessary precautionary measures can be taken without delay. 
Alphabetic filing is not necessary as cards can be quickly removed from any 
position in the pack* • 4 . 

A punched-card system codifying some basic characteristics of radio¬ 
isotopes has been used by the Western Division of Tracerlab., Inc., Rich¬ 
mond, California. The purpose of this card catalog is to facilitate the 
identification of unknown radioactive isotopes on the basis of half-life and 
radiation and to eliminate tedious searching of tables. It identifies isotopes 
also by their modes of formation, percentage of total radiation, decay 
schemes, and conversion coefficients. By eliminating stable isotopes and 
those with half-lives of less than five hours, the number of cards was 
reduced from 1000 to 375. The system uses McBee Keysort cards. The 
code chosen involves fourteen simple characteristics which are assigned 
but one hole each. These simple properties are: emits alpha particles, 
emits negatrons, emits positrons, emits gamma radiation, emits electrons, 
decays by K capture, decays by isomeric transition, is a fission product, 
has a radioactive daughter, belongs to the thorium series, belongs to the 
neptunium series, belongs to the uranium series, belongs to the actinium 
series and is naturally occurring. Seven quantitative properties are also 
coded, such as half-life, maximum gamma, positron, negatron and alpha 

*Wachtel, Irma S., “Indexing Nuclear Data on Punched Cards. Preliminary 
Edition,” USAEC Technical Information Service, TID-469, (April 26,1961); Appendix 
on page 7. 

* Brochure from Radioactive Products, Inc., 443 West Congress, Detroit 26, Mich. 
“Rapid Nuclide Identification and Selection by 9 Major Classifications.” 

4 Bonino, J. J., and Laing, K. M., “Punched-Card Classification of the Nuclides,” 
Nucleonics, 10, (2), 68, (Feb. 1953). 



REVIEW OF APPLICATIONS 


305 


energy, and predominant gamma and negatron energy. Each of these 
properties requires a group of four holes to describe the approximate half- 
life and energies of a given isotope. In addition, a group of eight holes is 
used to designate the atomic number. Only slightly more than one-half of 
the available holes are now assigned, assuring the continued and expanding 
usefulness of the cards 6 . 

Biological Data 

The applications of punched cards are perhaps more numerous and varied 
in the field of biological data than in any other. This may be due to the 
fact that there are a larger number of variables in this field, most of which 
cannot be controlled as are temperature, pressure and concentration in 
chemistry and engineering. 

The Dow Chemical Co. at Midland, Michigan, is using IBM cards to 
facilitate coordination between the work of the chemist and the biologist. 
A combination of numerical and alphabetical coding is used to identify the 
results of 75 different test procedures on 11,000 compounds. At the present 
time, approximately 200,000 punched cards are required to record the 
results of these tests. The system described has combined many of the 
functions of a biological clearinghouse with routine reporting of current 
biological tests. The use of nontechnical personnel to handle the bulk of 
the reporting, correlation, and indexing has greatly decreased the amount 
of time spent by professional research personnel on clerical work. The 
primary organization is by chemical entity, using the Chemical Abstracts 
names, a chronological serial number, and a structural classification num¬ 
ber. Numerical codes are used for the scientific name of each test, the test 
method, concentration of the chemical tested, and the biological result of 
the test 6 . 

A bibliography of 5,000 references is maintained at the Roscoe B. 
Jackson Memorial Laboratory, Bar Harbor, Maine, of all papers on specific 
inbred strains of mice, named genes in mice, or named transplantable 
tumors. The references have been classified on 5 x 8 inch Keysort cards 
(Figure 14-1) and are separated into periods of years (1930-34, 1935-39, 
etc.). They may be sorted directly by subject or strain individually named 
in the margins of the card. On the other hand, the cards may be sorted 
indirectly for which it is necessary to consult a key or index and needle a 

* Luke ns, H. R., Jr., Anderson, E. E. and Beaufait, L. J., Jr., “Punched Card 
System for Radioisotopes,” Anal. Chem., 26, 651 (April 1954). 

• Dunn, E. E. and Lynn, G. E., “Reporting and Indexing Biological Data by IBM 
Punched Card Methods,” presented at the American Chemical Society Meeting, 
March, 1952. 



306 


PUNCHED CARDS 



Figure 14-1. Keysort cards used for bibliographic references at the Roscoe B 
Jackson Memorial Laboratory. 

code number to locate references to the branch of a field of interest or a 
given minor strain or the named transplantable tumor sought 7 . 

A punched-card system was started at Lilly Research Laboratories in 
Indianapolis to separate for comparison purposes antibiotics with specific 
groups of properties. The file was originally established as a name file to 
supplement Baron’s “Handbook of Antibiotics.” A standard 80-column 
IBM card is used which contains no special printing and on which are 
punched 15 classifications of physicochemical data, 5 classifications of 
biological data in vitro, 8 classifications of biological data in vivo and 5 
classifications of biological data in vivo-toxicity. Definite information 
references supporting the groups punched on the IBM cards are contained 
on master cards which are different colors to enable quick location in the 
files. The master cards are related to the proper IBM cards through the 
corresponding file number. The user who wants information on a specific 
antibiotic goes directly to the master card file which is arranged alphabetic- 
ally. The coding outline makes use of positive results only and the file is 
intended as a guide to the literature, not as a substitute for it. The primary 
emphasis of the file is chemical but a limited basis for biological comparison 
is also provided 8 . 

Dr. Saul M. Bien, Lynbrook, New York, has devised a system for 

7 Staats, J., “A Classified Bibliography of Inbred Strains of Mice,” Science, 119, 
(3087) 295-296 (Feb. 26, 1954). 

* Ohrmund, Margaret, “An Antibiotic Literature File for Chemists,” presented 
before the Symposium on Pharmaceutical and Medicinal Literature, Division of 
Chemical Literature, American Chemical Society, Sept. 16, 1954. 







REVIEW OF APPLICATIONS 


307 



Figure 14-2. Keysort cards used for registering orthodontic diagnostic data by 
Dr. Saul M. Bien. 


registering orthodontic diagnostic data on hand-punched and hand-sorted 
McBee Keysort cards. (Figure 14-2). A total of 305 different items ab¬ 
stracted from the patient’s history, physical examination, radiographs, 
photographs and models may be punched on the card in code. The teeth 
are numbered according to the universal system, starting with the upper 
right as number one and ending with the lower right as number 32. A 
punched position on the cards indicates the absence of the tooth assigned 
that number. Cephalometric and other anthropometric data can be re¬ 
corded on the face or back of the card 9 . 

Bucknell University has established a file of edge-notched cards which 
constitutes a bibliography of references to studies published on the golden 
hamster. The cards have been prepared so that authors may be arranged 
in alphabetical order, or cards may be selected according to journal, date of 
publication, author and subject. References on bacterial diseases, cancer, 

• Bien, Saul M., “Registration of Orthodontic Diagnostic Records for Statistical 
Evaluation,” Am. J. Orthodontics, 41, (6), 482—183 (June 1955). 











































308 


PUNCHED CARDS 


caries, parasitology, pathology, and virus diseases account for most of 
the total publications. The file has proved very useful in the investigation 
of the genetics of the hamster, and in pointing out that very little is known 
about the requirements for optimal growth and reproduction, or anatomy 
of the hamster 10 . 

A description of a coding system using 3x5 inch punched cards has been 
reported by Norman D. Levine, College of Veterinary Medicine and Agri¬ 
cultural Experiment Station, University of Illinois. This system has been 
used to set up a file of several thousand cards for handling abstracts of 
veterinary, medical and general parasitology using issues of Biological 
Abstracts as a basic source. A decimal code covering three fields of four 
holes each is used to code the most important parasite genus discussed in 
the paper. This permits the use of 999 numbers in the code. A second set of 
three fields is used for a second parasite genus. A decimal code covering 
two fields and allowing for 99 numbers is used for the last genus. The first 
and second subjects are assigned five holes each and cover such items as 
cultivation, diagnosis, evolution, excretion, genetics, growth, regeneration, 
etc. An additive code based on five holes is used for the first letter of the 
first author’s name. Finally, three single holes are punched separately 
when the paper discusses more than two parasite genera, more than one 
host, or more than two subjects, respectively. This leaves a group of seven 
holes which are available for other information one might wish to code 11 . 

A systematic review of the world literature on vision in invertebrate 
animals was begun in 1946 by Lorus and Margery Milne, University of 
New Hampshire, on hand-notched and hand-sorted 3)4 x 7)4 inch Keysort 
cards. At present, abstracts have been typed on more than 4,500 of these 
cards which have been coded according to a system devised by the authors. 
The card file is arranged so that it can be sorted alphabetically by author 
and then chronologically under each author’s name, by subject and chrono¬ 
logically under each subject, and by taxonomic group. Twenty holes are 
assigned to code the first four letters of the senior author’s name. Addi¬ 
tional cards without abstracts are prepared for junior authors with a cross 
reference to the senior author’s card. Nine holes are used to code the date 
of the reference; two fields of 7, 4, 2, 1 to cover the units and decades, and 
a ninth hole to be notched for any date earlier than 1900. Subject categories 
were punched in sixteen holes and coded according to the “Classification 
of Zoological Literature” used by the Wistar Institute Bibliographic Serv¬ 
ice. A single hole is punched to indicate that an abstract has been made 

10 Magalhaes, Hulda, “The Golden Hamster as a Laboratory Animal,” J. Animal 
Technicians Association, 5, (2), 39-44 (September 1954). 

11 Levine, Norman D., “A Punched Card System for Filing Parasitological Bibliog¬ 
raphy Cards,” J. of Parasitol., 41, (4) 343-352, (August 1955). 



REVIEW OF APPLICATIONS 


309 


and the reference checked for accuracy. Thirty-three holes remain for 
taxonomic use, twenty of which are used for the first four letters of the 
generic name. Dewey’s system of decimal classification was followed to 
code for taxonomic position, the numbers being abbreviated by deleting 
the initial (common) 59. Thus the Dewey 593.1 for Protozoa became 3.1. 
Some material was typed on the face of the card: authorship, title in the 
original language, journal reference as it appeared on the title page of the 
volume rather than in the form used in the Union List of Serials, the ab¬ 
stract and the source of the reference. The code for the source of the 
information was a simple one. For example, 24 Brown 627 would indicate 
that the reference was cited on page 627 of a paper written by Brown in 
1924. 

From the standpoint of cumulative experience with this system, the 
authors have several suggestions on ways in which it could be improved 
(Figures 14-3 and 14-4). For example, E-Z Sort cards with six holes per 
inch could be used instead of the Keysort which have only four per inch. 
Thus the same size card would accommodate 50 per cent more holes and 

Notched •!(<)• depth - microfilm of 



Figure 14-3. Keysort card used for controlling literature on vision in invertebrate 
animals. 


Wi 



310 


PUNCHED CARDS 



Figure 14-4. Kevsort card used for cont rolling literature on vision in invertebrate 
animals. 


the percentage assigned to author (now 25 per cent) would fall sharply. 
If a card with two rows of holes were used it would mean a further saving 
in space, but any name having the same letter in two or more of the first 
four positions could introduce confusion. Another E-Z Sort card with four 
rows of holes could be used, notching the initial letter of the author’s 
name single depth, the second letter double depth, and so on. Different 
systems have been suggested for expanding the alphabet to 30 characters 
or simplifying it to 23 characters. Another improvement discussed was 
the possible use of ten holes for the date in order to indicate the century 
more definitely. This would mean an increase of two holes, the first of 
which would represent the 20th century if uncut and the 19th if cut, 
while the second hole would indicate the 18th century if cut and the 17th 
if uncut. An additional twelve holes would improve the manner of coding 
the journal reference using a three-letter designation with double-row’ 
perforations. For example, JAB would stand for the Journal of Animal 
Behavior, JGP for the Journal of General Physiology, etc. Another hole 
to indicate that a copy of the reference is in the file and one to show that 
a microfilm copy is in the file would also add to the usefulness of the file. 
The Dewey Decimal system proved to be of little value in this applica- 

















REVIEW OF APPLICATIONS 


311 


tion. Also the use of twenty holes for the generic name was wasteful of 
space. The number of papers discussing a variety of organisms grew 
beyond expectations. For these, only direct coding would have been help¬ 
ful. It is estimated that a list of 55 phyla, subphyla, classes and subclasses 
would handle all of the great variety of animals upon which photosensory 
studies have been published 12 . 

IBM punched-card methods have been perfected by the Chemical- 
Biological Coordination Center, at the National Research Council in 
Washington, for coding information concerning the biochemical trans¬ 
formations undergone by pesticides in the course of their metabolism. A 
code based upon a comprehensive classification of enzymes permits the 
recording of the effects of pesticides upon these catalysts. This information 
can be retrieved by searching for the type of reaction, the organ, the 
species, the names of the pesticides, and their products. An abstract 
suitable for coding is prepared on a code sheet form. The information is 
broken down into coding fields, such as taxonomy, organ, host, specific 
effect, general effect and dose level. Punched cards are then prepared 13 . 

Photography 

A bibliography on photographic theory originally prepared at Eastman 
Kodak has been established on McBee Keysort punched cards. The 
bibliography covers emulsion making, latent image formation and develop¬ 
ment. Color photography and sensitometry are not included, except to a 
minor extent. The card used measures 634 x 734 inches and is specially 
printed to indicate holes for coding the senior author’s name, the date and 
type of publication and the information mentioned above 14 . 

Keysort punched cards have also been used by the Edwal Laboratories, 
Inc., Ringwood, Illinois, for a file of photographic references in which 
abstracts are pasted to the cards. The file may be searched by main sub¬ 
ject, author’s name, date of publication or patents. Main subject coding 
is numerical according to the Universal Decimal System as used by both 
Kodak Abstracts and Photographic Abstracts, the two abstract journals 

M Milne, Lorus J., and Margery, “Foresight and Hindsight on a Punch-Card 
Bibliography,” report on work done at the University of New Hampshire, par¬ 
tially supported by the graduate school and at the Scripps Institution of Ocean¬ 
ography, University of California. Contribution from the Scripps Institution of 
Oceanography, New Series, No. 968. 

13 Wood, G. C., and Welt, I. D., “A Multi-indexed Machine Sorted, Punch Card 
System for Pesticide Metabolism Data,” Agriculture and Food Chemistry , 4 (10), 
886-888 (Oct. 1956). 

14 LuValle, James E. Item #22 in “Abstracts of Presentations. Investigators 
Restricted Seminar #1 on the Chemistry of Photographic Processes,” Chicago, 
Illinois (Sept. 4, 1953) Sponsored by the Chemical Division of the Headquarters Air 
Research and Development Command. 



312 


PUNCHED CARDS 


used as the main source for the reference file. The alphabetical coding 
of the author names is according to the revised method of Cox, Bailey and 
Casey, Chemical and Engineering News, September 25, 1945. Another 
suggested use for these cards is as an index to Kodachrome medical slides. 
These can be filed according to pathological condition, part of the body 
affected, symptoms, etc. 16 . 

Punched cards have also come into use in the filing of photographic 
negatives. E-Z file cards with Filmsort positive transparencies of the 
print have been used by Eastman Kodak for this purpose. Different 
classification systems are necessary to suit the type of work being done 
by individual photographic organizations. For the commercial photog¬ 
rapher, the name of the subject or client serves as the best filing key. 
This is true also for such files used for police identification, and by industrial, 
personnel passport and medical photographers. Commercial illustrators 
and industrial photographers, on the other hand, are more concerned 
with jobs or products and for them filing by product name, company or 
department, job name or job number is more desirable. Coded punching 
around the edge of the card carries the key to the filing classification for 
the photograph covered by each card and allows rapid mechanical selec¬ 
tion of the desired cards. Print and data sheet transparencies may be 
examined directly from the card with a viewer without removing the 
negative from the file. This system fulfills the two most important req¬ 
uisites for filing negatives: prompt location and minimum handling to 
avoid scratching the surface and damage from dust particles 18 . 

Laboratory Records 

One of the most important factors in the successful operation of an 
industrial analytical laboratory is an adequate system of keeping records 
of samples being analyzed and of the data obtained. A system described 
by A. H. Hale and J. W. Stillman of the E. I. du Pont de Nemours & Co., 
Inc., in Wilmington, has been developed to meet this need and is in actual 
use in several industrial laboratories. Printed slips with interleaved car¬ 
bons are used for recording and reporting analytical data. From these 
slips punched cards are prepared, first to provide positive control of the 
progress of the analysis and then to locate the original data on the report 
slips in the permanent file. McBee Keysort cards (5x8 inch) are punched 
to show the assignment of samples to be analyzed, the progress of the 
analyses, the location of the analytical report in the files and to provide 
information on the operation and efficiency of the laboratory. Code num- 

“Hill, Thomas T., “Finding Photographic Information,” J. Biol. Photographic 
\s8ociation y 17, (3) 103-114, (March 1949). 

16 “Filing Negatives and Transparencies”, a twenty-page pamphlet prepared by 
Eastman Kodak Company, Rochester, New York, (Oct. 1953). 



REVIEW OF APPLICATIONS 


313 


bers are used to designate each chemist so that it is possible to determine 
the name of the worker from the number punched in the upper right 
comer of the card. The year and the month of receipt of the sample are 
punched in the upper left comer. The material block has space for 4,000 
numbers to include all kinds of samples received. Five-hundred jjpmbers 
are assigned to organic compounds and 700 to inorganic compounds. 
There are sections to notch for polymers and their modifiers (colors, 
fillers, inhibitors, etc.) 17 . 

Marginally-punched cards are used by the Strong Cobb Company of 
Cleveland, Ohio, as part of the record-keeping system for its pharmaceuti¬ 
cal control laboratories. A 6 x 8 inch Key sort card serves as the permanent 
record of all analytical control data for each product manufactured. The 
card is custom printed to specification thus eliminating the unnecessary 
copying of information which was formerly transcribed by hand for each 
job. Space is provided on the printed portion of the card for the date of 
receipt of the sample, the stage of production, assay data, etc. Information 
is coded for all constituents assayed, manufacturing difficulties encountered, 
and the type of product being developed. A similar system is used by the 
White Laboratories of New Jersey 18 • 1# . 

As long as ten years ago, workers at Eastman Kodak Co. in Rochester 
realized that some sort of a punched-card system was essential to locate 
more quickly the approximately 1,000 organic compounds that had been 
synthesized in the laboratory for research purposes. It was preferable to 
locate classes of compounds rather than individual substances. Different 
colored 5x8 inch Keysort cards (Figure 14-5) were printed to certain 
specifications; each color represents one broad group of organic compounds. 
These cards are filed separately according to colors. Structural groups 
and features are listed in two rows along the top of the card, together with 
a few types of substances of special interest such as ureas. Along the 
right-hand edge are numbers referring to ranges of light absorption, while 
those on the left refer to numbers and positions of substituents in simple 
cases. Halogens and amines (primary, secondary and tertiary) are also 
included on this side. The bottom is reserved for ring sizes, some groups 
of substances, a few less frequently encountered structural features and 
the very important section, labeled “hetero.” No provision has been made 
for salts (anions) since this was unimportant in the present application. 
The order of precedence of groups (and so, of card colors) was arbitrarily 

17 Hale, A. H., and Stillman, J. W., “Development of an Efficient Analytical 
Record System,” Anal. Chem., 24 (1) 143-149 (Jan. 1952). 

11 Naimark, G. M., and Prindle, R. F., “Pharmaceutical Control Laboratory 
Record System,” Anal. Chem., 26 (4) 645-647 (Apr. 1954) 

'• Naimark, G. M. “Industrial Analytical Record Keeping,” Drug and Cosmetic 
Industry (Sept. 1955). 



PUNCHED CARDS 


314 



Figure 14-5. Key sort card used to record information on organic chemical com¬ 
pounds at Eastman Kodak Co. 


set: dyes, heterocyclics, aromatics, cycloaliphatic and aliphatic, in that 
order. For example, phenyleicosane is put on an aromatic card even though 
the phenyl group is the smaller part of the molecule. The structural formula 
is written in the blank spaces on the card. About 9,000 cards can be sorted 
per hour using the Keysort Selector. With a very few modifications, this 
same card could be used by organic chemists in general 20 . 

Technical Services 

A group of chemical companies have been developing mechanical methods 
of product application data handling. Methods used by Shell Chemical Co. 
and others offer speedier literature searching and increased ease in famil¬ 
iarizing new personnel with the previous work. Shell Chemical Co. breaks 
down the information in laboratory reports into application, composition, 
and properties. Each of these is assigned one edge of an edge-punched 
card. Each major group has sub-groups such as graphic arts, photographic 
processes, the presence or absence of ketones, esters, alcohols, diluents, 
solvency, etc. Each card also contains space for the written recording of 
pertinent data such as references to the original report and the test data 
(solvency, dilution, density, etc.). Shell Chemical’s solvents group finds 
that an average of one to ten punched cards is necessary to code the 

10 Allen, C. F. H., “Keysort Punch Card System As Used in the Organic and 
Polymer Chemistry Department, Chemistry Division, Eastman Kodak Company.” 
April 25, 1956. Report submitted to J. W. Perry by the author. May 11, 1956. 












REVIEW OF APPLICATIONS 


315 


ordinary laboratory report. Fifteen years of records are now coded into 
5,000 cards. Within 30 minutes reference to all of the work previously 
done in a specified area can be located 21 . 

Micro Switch, Freeport, Illinois, has initiated a project to determine 
how product histories, or abstracts of company “know-how,” should be 
prepared and what would be their long-range value. One of the major 
tasks in the preparation of these histories is accumulation of the informa¬ 
tion needed. A ready means of retrieving all the available data on such 
products is essential. As bits of knowledge come in from such divisions as 
Products Research, Engineering, Methods, and Quality Control, they are 
recorded on specially designed punched cards (Figure 14-6). The proper 
classification, coding and punching of the cards make it a relatively simple 
matter to sort out all information on a given product, on any given process, 
on quality control data and types of materials. This eliminates duplication 
of files and cross indexing 22 . The author’s name code is based on a 100- 
division alphabet breakdown and is punched in two fields on the left side 
of the card. At the upper right two fields are devoted to coding the prod¬ 
uct group. One field gives the type of switch involved and the other the 
catalog listing. The various parts of the switches such as springs, anchors, 
or plungers are punched into two fields at the bottom. In the two fields 
next to this, are coded all processes (such as welding or heat treating). 
Another two fields at the bottom are designated for coding materials 
involved. Specifications (resistance, ductility, etc.) and a miscellaneous 
classification use two fields each on the right side. All items in each classifi¬ 
cation are numbered consecutively from one. The punched card carries 
the identifying file number for locating the original information. 

Market Research 

An edge-punched card system was adopted by Magnaflux Corporation 
of Chicago as a logical approach to the task of storing and retrieving 
details on current records of customers and prospective customers. Specially 
designed 5x8 inch E-Z Sort cards with 140 holes are used (Figure 14-7). 
Both sides of the card are printed; white cards are used for customer 
information and buff cards for prospects, and both of these are filed sepa¬ 
rately. There are approximately 1,200 customer cards—300 active and 
900 inactive. The field in the upper right-hand comer is used to code the 
first two letters of the company name. Letters A through Me are assigned 
numbers one through thirteen. Letters M through Z are given numbers 
V, VI, V2 etc. through V13. A four-digit number is coded in standard 

*' “Punch Cards Up Tech Service Productivity,” Chemical Week, (Nov. 3, 1956). 

" Rhynders, R. W., “Product History Clears Haze from Technical Records,” 
Industrial Laboratories, (Feb. 1956). 




Figure 14-6. Keysort cards used for recording information from company reports 
at Micro Switch. 


316 


























REVIEW OF APPLICATIONS 


317 



itiuifuiwiw nui 



v 

r 



Figure 14-7. Card for recording information on customers at Magnaflux Corpora¬ 
tion. 


six-hole fields at the top center of the card to indicate the name of a com¬ 
pany in a given area of business. The business or individual is identified 
in a numerical method by the product manufactured or service rendered 
using the Standard Industrial Classification. The field on the right-hand 
side of the card is devoted to the inspection methods marketed by the 
corporation. Specific equipment is represented in the next five fields on 















318 


PUNCHED CARDS 





Figure 14-8. Card containing microfilm copy of customer correspondence, as used 
at Magnaflux Corporation. 

the bottom and lower left-hand corner of the card. Each correspondence 
file is transferred to the individual customer card by microfilming the 
documents on adhesive-backed film (Figure 14-8). In this way all of the 
documents pertaining to the customer’s file can be scanned by turning 
over the punched card 23 . 

Meteorological Data 

IBM punched cards have been used in handling large masses of experi¬ 
mental data resulting from upper atmosphere research. Hand-sorted 
cards could have been used but their capacity is comparatively limited. 
IBM bibliography cards may be selected by author’s name, publishing 
agency or journal name, date of publication, security classification, lan¬ 
guage or subject matter. Abstracts of the articles are typed on the back of 
the card. A recommendation by Boston University’s research groups was 
made for the establishment of a central agency to abstract, index and 
distribute the cards 24 . 

In 1948 the British Meteorological Office began to index upper air 
data using Hollerith card systems. Observations were punched on the 

23 Cannon, W. A., Jr., “A Punched Card System for Technical Liaison, Sales 
Analysis, and File Reductions,” Hathaway Instrument Division, Hamilton Watch 
Co., Chicago, Illinois. 

24 Low, Ward C., “Technical Publication Abstracts on IBM Punched Cards I,” 
Technical Note ft 15, July 14, 1952. Upper Atmosphere Research Laboratory, Boston 
University. 










REVIEW OF APPLICATIONS 


319 


cards at the observation stations and then sent to the central office for 
machine sorting and tabulation. It was anticipated that a little more than 
130,000 cards would be prepared each year to make available all statistical 
information about upper air conditions over the British Isles. Coded 
information includes date, time, place, pressure levels, wind direction, 
temperature and humidity 25 . 

Hollerith cards were adapted in 1950 to evolve linear-function tables 
for upper air data. Owing to the large range of values and the varying 
number of observations, 40,000 totals had to be given to cover the ranges 
100° to — 100°F and 31 to 11 observations. It is doubtful that the prepara¬ 
tion of such tables, as invaluable as they are, would have been worth 
while by clerical workers as too much time would be lost in the mechanics 
of working out monthly mean values 2 *. 

In 1951 Hollerith cards were used to great advantage in the field of 
marine meteorology and they made possible the preparation and publica¬ 
tion of climatological atlases of the oceans. After various meteorological 
elements included in the observations were punched in appropriate codes, 
the cards were sorted and filed in packs according to the month and the 10 ° 
Marsden Square in which they belong. A total of approximately 3 j-£ 
million British and 6^2 million German cards were filed in the Marine 
Branch at that time. The effect of wind velocity on the sea and air tempera¬ 
ture, the diurnal variation of the sea and air temperatures in relation to 
cloud amount and to the sea and air temperature differences, and the 
effect of wind velocity on relative humidity were investigated 27 ■ **. 

Geological Data 

Geologists, like other scientists, have been losing ground in their efforts 
to keep informed on latest developments in their realm of specialized 
knowledge. In order to help alleviate this situation the Petroleum Re¬ 
search Corporation of Denver, Colorado, has developed and is producing 
the Micro-Research-Card Library of the Rocky Mountain region. This 
reference collection is available for purchase or rental. The library contains 
microphotographic reproductions of approximately 8000 published and 
unpublished articles and theses. Included are papers from more than 80 
periodicals, publications and unpublished reports of the U. S. Geological 

** DeWar, D. “The Hollerith Card System Applied to Upper Air Data,” Meleorol. 
Mag., 78, 163-166, (1949). 

** DeWar, D. “Preparation of Linear-Function Tables on a Hollerith Tabulating 
Machine,” Meteorol. Mag., 79, 137-140 (1950). 

17 Gordon, A. H., “Development of Modern Techniques in Marine Meteorology,” 
Meteorol. Mag., 80, 78-83, (1951). 

75 Gordon, A. H. “Adaptation of Mechanical Sorting and Tabulating Machines to 
Research in Marine Meteorology,” Meteorol. Mag., 80, 269-270, (1951). 



320 


PUNCHED CARDS 


Survey and an almost untapped wealth of theses. The material in this 
file is so arranged that it can be sorted, in a matter of a minute or so, by 
geologic subject, by area, or by geologic time. Any given article can also 
be selected by author. The file now contains about 8000 5 x 8 inch film 
transparencies, perforated at one end with 207 small holes. Each hole 
represents a geologic subject, a geologic time, date of writing, or geographic 
area within the Rocky Mountains. Every article has been coded by a 
geologist whose notations appear on the card in microfilm form and also as 
slots extended from the holes in the punched end of the card. Cards are 
needle sorted. The number of categories needed for optimum separation 
was determined by the subjects and areas covered. Each card is numbered 
and filed according to the area its subject encompasses. Numbered areas 
within the greater Rocky Mountain region are indicated on a map in the 
library. In addition to the numbered breakdown by areas, letter designa¬ 
tions (N, S, E, W, and C, for north, south, east, west and central) indicate 
the portion of a numbered area involved. To facilitate selection by area 
further, a separate coding is made by states. Eighty-seven geologic sub¬ 
jects are coded, including structural contour maps, paleontology, radio¬ 
active minerals, origin, oil analysis, etc. If a given article is sought, a 
printed bibliography fists, by author, all articles included in the library 
and gives the “call number” which is slotted on each card. The first two 
digits of the call number are the area designations and these are slotted 
along one edge of the card so that a card out of position in the file is detected 
at once. Based on the demand, as indicated by the first effort, the library 
may be expanded to coverage of the continent, and even of the world**. 

The accumulation of data resulting from a program undertaken by the 
Ohio Division of Geological Survey to evaluate the coal reserves of the 
state bed by bed has been compiled on punched cards. The system of 
tabulation selected utilized IBM equipment (a key punch machine for 
punching data into the cards, a sorter which mechanically sorts the cards 
into any desired order, and a tabulating machine which mechanically 
prints the data from the punched cards). The file contains approximately 
12,000 individual outcrop records from the 25 coal-bearing counties of the 
state. The tabulation procedure was designed first to study the problems 
in the Pennsylvania part of the Ohio geologic section. The transfer of the 
information from the original file sources to punched cards began in 1954. 
For the 14 counties evaluated so far, over 4,000 stratigraphic records have 
been coded containing more than 10,000 observations of individual coal 
beds and their associated strata. The code as set up uses a four-digit 
stratigraphic number for each coal bed and a one-digit lithologic number 
within a single coal-to-coal interval. The lithologic number consists of a 

** Chronic, John, “How Microfilm Library Aids Research,” World Oil, (May 1956). 



REVIEW OF APPLICATIONS 


321 


one-digit number for each of the nine more common rock types such as 
marine limestone, sandstone, etc., which commonly occur in the interval 
between coal beds. The four-digit stratigraphic code serves to designate 
geologic age and position within the standard geologic column of Ohio. 
The one-digit number code serves to designate the lithology and position 
of the strata within the depositional cycle comprising the interval from 
the base of one coal bed upward to the base of the next younger coal bed. 
Counties are coded numerically from 01 to 88 in accordance with their 
alphabetic position and townships are coded numerically in accordance 
with their alphabetic position within the county. Every digit of the standard 
80-column IBM card is used in the general part of the study, necessitating 
the use of a second card for tabulation of interval data. Identification 
data such as file number, sources and location are repeated for each cycle, 
so that reference to original sources may be made at any point regardless 
of the subsequent classifications in which any one cycle appears. A second 
card is punched for each interval in the section. A competent operator 
can assemble data at the rate of 700 to 1000 cards per day. 30 

Astronomical Data 

Lick Observatory at the University of California is engaged in the 
preparation of a list of all double star measures beginning with the year 
1927.0, the closing date of Aitken’s “New General Catalog of Double 
Stars.” All measures for double stars of the Northern Hemisphere that 
have appeared in print or in manuscripts have been entered on more than 
80,000 IBM punched cards. The cards are proving so useful and efficient 
for compiling lists and carrying out statistical studies that the measures 
cataloged for many years at the Southern Hemisphere are being sent to 
the Observatory from the Union Observatory in South Africa, to be 
punched. Other files maintained at Lick Observatory include punched 
cards for some eclipsing binaries. New projects for computing are being 
planned which by former methods would take more than an astronomer’s 
lifetime. Other applications in the field of astronomy include the extended 
moon and planetary ephemerides that have been constructed by Brouwer, 
Eckert and Clemence, the automatic plate measuring machine developed 
at the Watson Scientific Computing Bureau by Eckert and his associates, 
and the use of punched-card processes to calculate the orbits and ephe¬ 
merides of comets and minor planets by Herget and Cunningham 31 . 

10 Smith, William A.; Brant, Russell A.; and Klein, Marian S., “An Application of 
Business Machine Technique to Stratigraphic and Coal Resources Studies,” Infor¬ 
mation Circular No. 18, State of Ohio Dept, of Nat. Resources, Div. of Geological 
Survey, 1956. 

11 Personal Communications: Feb. 24, 1956, Mrs. James F. Chappell, Lick Ob¬ 
servatory to Robert S. Casey and March 23, 1956, H. M. Jeffers, Lick Observatory 
to Robert S. Casey. 



322 


PUNCHED CARDS 


Legal Data 

A special committee of the New Jersey State Bar Association made an 
investigation into the realm of mechanized and automatic literature 
searching, an explanation and resume of which was published in 1953**. 
This investigation showed that a crisis existed in the storage and use of 
legal literature to a degree which demanded an immediate solution whereby 
the material read by human eyes and brains could be examined and selected 
at a rate at least 100 times greater than that of which humans are capable. 
In his article, Biunno suggests a guide for preliminary experimentation 
and dicusses the use of a machine language called “Luko” in which 20 
consonants (Q being excluded) and 5 vowels are arranged in all possible 
combinations to provide code headings. 

The American Bar Foundation of Chicago, Illinois, conducted an experi¬ 
ment designed to demonstrate one possible adaptation of punched cards 
and mechanical retrieval to legal research processes as performed daily 
throughout the country by lawyers and judges. Remington Rand cards 
and equipment were used to demonstrate this system at the Symposium 
on Systems for Information Retrieval, held in Cleveland in April 1957. The 
Illinois Divorce Statute was used as the master code and the cards were 
punched on the theory that the Illinois lawyer would want to know only 
to what extent the Idaho law, for example, is different from the Illinois 
law with which he is most familiar. A sheet was prepared showing the 
correlation between the holes in their numbered positions on the cards and 
the master code. The American Bar Foundation also hopes to use punched 
cards and the Remington Rand installation in the membership department 
of the American Bar Association to index the publications of professional 
legal organizations which are not now covered in the Index to Legal 
Periodicals® 3 . 

A suggested coding method for legal data using a specific area as a test 
case (liability of electric power and telephone companies for injury or 
damage by lightning transmitted on wires) has been received from Charles 
Cobb, Jr. 34 . The plan would use 5x8 inch E-Z Sort cards with two rows of 
holes along each edge (Figure 14-9). An analysis suitable for all purposes 
would require the use of a logical scheme adapted to the expression of the 

** Biunno, Vincent P., “Searching Legal Literature—An Appraisal of New Meth¬ 
ods,” Law Library Journal, 46, (2), 110-119, (May 1953). 

** MacKinnon, F. B.; Leary, J. C.; and Levinson, D., Jr., “An Analysis of the 
Problem and An Experimental Adaptation of Punched Cards and Mechanical Re¬ 
trieval to Legal Research and Indexing of State Statutes, Codes and Session Laws,” 
Prepared for the Symposium on Systems for Information Retrieval, April 15-16, 
1957, Cleveland, Ohio. 

* 4 Personal Communication, June 20, 1956 from Charles K. Cobb, Jr., Law Book 
Department of Little, Brown and Company, Boston, Mass, to James W. Perry. 



REVIEW OF APPLICATIONS 


323 



Figure 14-9. Card used for recording legal data by Charles Cobb, Jr. 


intentions of legal propositions. Four numerical fields would be used to 
describe the defendant (in this case one number for electric power com¬ 
panies and another for telephone companies), the plaintiff (a customer or 
member of the general public), the means by which the injury or damage 
was inflicted, and the nature of the injury or damage. The cases gathered 
from the legal digests and other sources are also sorted by state and are 
arranged chronologically for each state. The system would be adequate for 
sorting cases by legal result, both for final judgment and for the conditions 
of liability. 

Chemical Literature 

The great bulk of existing chemical literature and the continued high 
rate of production have greatly accelerated the search for some method 
or device to permit a rapid and thorough search of the whole breadth of 
this literature to verify the presence or absence of a given fact. A suggested 
system has been described in the literature 35 which combines a microcard 
with a punched-eard coding and sorting system. A standard 3x5 inch 
microcard is fitted into an area covering 60 columns of an 80-column IBM 
punched card. The remaining 20 columns are available for coding. Ten 
columns are punched with a standard Dewey-Decimal or other similar 
classification of the main subject matter or title of the paper reproduced 
on the microcard. The remaining ten columns are punched with a set of 

J ‘ Williams, T. J. and Rose, A., “A Solution to the Problem of Storage and Avail¬ 
ability of Chemical Literature,” J. Chern. Educ., 29, 146-147, (March 1952). 















324 


PUNCHED CARDS 


•f 

V71 

fl« 

•tl 

T lfiFWF? **••••••••• 

fj 

Clark, Ualtar E, and M. L. Holt 

El*etrod»position of eotelt-tuigsten alleys from a citrate bath 

• • 

ae 


J. Ha*trocban. Coe. 34, 244-52 (1143) 

*• 

•:| 

C.JU it, 45 (13*9) 

mm 

••• 

An aq. bath suitable for the electrodeposition of Co^W alleys, oontg . 

ag 

#t| 

appro*. 50t V, Is described. The both contains Co sulfate, Ha tteifstate, sod 
citric acid in the approx, sol ratio of lsltl.5. a bath pH of about 7 is 


til 

■ost satisfactory and is obtained ty the eddn. of W^OH. The bath teqp, 
should be 70 or above. A cathode c.d. of 15 aiap/sq. da. is suitable, 
although lover c.ds. decrease the V content of the deposit only slightly sad 

sie 

til 

•»• 

tit 

A 

result in higher current efficiencies. Bright cathode deposits are obtained 
over a vide ranee of c.ds. Anodes of Co or of V or of both of these oetalj, 
or of sore inert neterUl can be used. 

a# 

ae 



W 

tit 

IKNM 

23 

•J3> 

• • 

NUMHnnnuii ^ n <* * •< »i «• » •• • • i % . « t 

3 


Figure 14-10. Keysort card used for recording chemical data. 

randomly spaced four-digit random numbers, which make it possible to 
obtain up to several million different combinations of spacing and numbers 
useful for indexing. 

Background data and information about experimental techniques rele¬ 
vant to the graduate research problem accumulate to such an extent that 
informal notes are no longer adequate to keep the material readily avail¬ 
able. Punched cards make possible multiple classifications of an abstract 
or even an individual item of information and the problem is then reduced 
to deciding upon a coding system which will permit filing for finding. The 
description that follows was set up by the author for his graduate research 16 . 
A 5 x 8 inch McBee Keysort card is used with approximately 4x7 inches 
on each of the two sides available for recording information (Figure 14-10). 
There are two rows of holes punched around the margin giving a total of 
182 holes for coding. It was decided to use the first letter of the senior 
author’s last name in the author code, using the holes along the upper 
margin at the left end marked A, G, and A. In the outer row A is 1, G is 3 
and A is 9 while in the inner row A stands for 2, G for 6 and A for 18. If, 
for example, the senior author’s last name begins with “C,” the hole 
for three is punched. A pencil notation on the front of the card for this 
would read G 0 where G stands for three because the subscript “0” indi¬ 
cates the outer row of holes. The subscript “1” would indicate the inner 
row of holes. Separate 3x5 inch cards are maintained, filed alphabetically 

*• Orr, C. H., “A Punched-Card System for Graduate Research,” J. Chem. Educ., 
30, 140-142 (March 1953). 












REVIEW OF APPLICATIONS 


325 


according to source which permits the assignment of a number to each 
source as it appears. The holes labeled B through H at the top of the card 
are reserved for the source code, allowing a total of over 2,200 sources 
to be coded if necessary. In the date code the first outer hole I 0 corresponds 
to 1800 and the first inner hole, Ii, to 1900. The remaining holes in the 
group J through N, provide spaces for the ternary code used for the author 
and source. The subject code numbers are assigned to the outer and inner 
row of holes on the bottom and left-hand side of the card, which are num¬ 
bered from 1 to 46. The major division referring to specific plating practice 
is located from numbers 31 to 36, outer and inner holes. The subtopics 
are acidity, addition agents, agitation, temperature, etc. The code numbers 
for the metals are assigned to the outer and inner rows of holes along 
the right-hand edge of the card labeled with the chemical symbols for 
some of the elements. The reference and an abstract are typed on the 
face of the card. 

Stanley Kirschner, Department of Chemistry, Wayne State University, 
Michigan, has described his system for coding and abstracting chemical 
literature. He uses a specially designed IBM card (Figure 14-11) which 
allows much of the important information to be coded directly by punched 
holes and provides space for the title and a brief abstract on the face of 
the card. Two initials and the senior author’s surname may be punched 
directly into columns 1-10 on the card using the IBM letter code, which 
allows coding any letter of the alphabet by means of a double punch in a 
single vertical column of numbers. The name of the journal in which the 
original reference appears is coded into columns 11-14 by means of a 
four letter abbreviation such as that devised by Bishop for the “Coden” 
system [C. Bishop, Am. Documentation , 4, 54 (1953)]. The volume number, 
page and year may be coded directly into columns 15-24 using the ap- 



Figure 14-11. Punched card used for recording codes pertaining to chemical litera¬ 
ture by Stanley Kirschner, Wayne State Univcristy. 



















326 


PUNCHED CARDS 


propriate numerals. Only the last two digits of the year are coded. The 
name of the journal in which the abstract appears may be coded into 
columns 25-28 in a manner analagous to coding the name of the journal 
in which the original reference appeared. The volume number is coded 
directly into columns 29-31. Columns 32-36 are used to indicate the 
column or page number for the abstract and column 37 may be used to 
show the location of the abstract on the page. The individual subjects 
are coded into columns 40-69. A maximum of 360 subjects may be coded 
in this section using a single or direct punch per subject. A direct punch 
subject classification code in inorganic chemistry has been worked out by 
the author (Figure 14-12) with about seventy subjects. Five new sub¬ 
jects (on the average) are added every year. In columns 70-80 the small 
numbers above the row of zeros (or sevens) represent the first digit of the 
atomic number and the numbers in the vertical columns represent the 
second digit 17 . 

The file discussed next was developed at the U. S. Geological Survey, 
Washington, for the maintenance of a bibliography in geochemistry 38 . 
Standard 6% x 7}4 inch McBee punched cards with five holes per inch 
and a double row of perforations around the entire perimeter are used. 
The section in the upper right-hand corner is designated to code the 
author’s name. The numerical breakdown permits the coding of only one 
author per card. The first and second letters of the name are broken down 
alphabetically into 99 subdivisions. Such classifications are available from 
the card manufacturers. The same number of chemical elements may be 
coded in a similar manner by punching the number corresponding to the 
atomic number of the element. Two fields of four holes numbered 7, 4, 2, 1 
each are used in the upper margin of the card to code the year of publica¬ 
tion. The right field is used for units and the left for tens. By limiting the 
use of each field to nine entries, it is possible to attain 99 entries in the two 
fields using a total of only eight holes. The hole marked “zero” is punched 
when 10, 20, 30, etc., are desired and also when 01, 02, 03, etc., are to be 
indicated. Simple numbers are distinguished from combined numbers by 
deep punching instead of shallow. The century is neglected inasmuch as 
most entries for this particular file are in the 20th century. The section 
marked “main element” at the top center of the card was established for 

87 Kirschner, S., “A Simple, Rapid System of Coding and Abstracting Chemical 
Literature Using Machine-Sorted Punched Cards,” presented at the Atlantic City 
meeting of the American Chemical Society, Division of Chemical Literature, Fall 
1956. 

38 Breger, J., “Design of Simple Punched Card Systems, with Reference to Geo¬ 
chemical Problems,” accepted for publication in Economic Geology. 



Col. 

Numbers 

No. 





1 

2 

3 4 

40 

Acids & Bases 

Monodentate 

Raman Spectros- 



Ligands 

copy 

41 

Analytical Proce- 

Magnetic Moments 

Reaction Mech- 


dures 


anisms 

42 

Apparatus & 

Mass Spectrometry 

Research Pro- 


Equipment 


posals 

43 

Bidentate Ligands 

Microscopy 

Resolution Pro- 



i 

cedures 

44 

Biography 

Microwave Spectros- 

Review Articles 



copy 


45 

Biological Chem- 

Nomenclature 

Rotatory Disper- 


istry 


sion 

46 

Bond Types & 

Non-aqueous Sol- 

Safety 


Properties 

vents 


47 

Chromatography 

Nuclear Chemistry 

Science Reviews 

48 

Cyclopentadienides 

Nuclear Magnetic 

Special Com- 



Resonance Spec. 

pounds 

49 

Dielectric Constants 

Organometallic Com- 

Stability Con- 



pounds 

stants 

50 

Dipole Moments 

Organosilicon Chem- 

Stereoisomerism 



istry 


51 

Coordination Com- 

Organozirconium 

Theory 


pounds 

Chemistry 


52 

Education 

Oxidation-Reduction 

The Trans Effect 

53 

Electrolysis (not 

Patents 

Ultraviolet Spec¬ 


polarog.) 


troscopy 

54 

Emission Spectros¬ 

Periodic System 

Unusual Oxida¬ 


copy 


tion States 

55 

Geometric Isomer¬ 

Photochemistry 

Visible Spectros¬ 


ism 


copy 

56 

Hexadentate Ligands 

Photometric Titra¬ 

X-ray Spectros¬ 



tions 

copy 

57 

History of Chemis- 

Polarography 

Zone Melting 


try 



58 

Industrial Chemis¬ 

Poly-acids 



try 



59 

Infrared Spectros¬ 

Polydentate Ligands 



copy 



60 

Ion Exchange 

Polymerization 


61 

Kinetics 

Preparations, Lab¬ 




oratory 


62 

Laboratory Tech¬ 

Properties, Chemi¬ 



niques 

cal 


63 

Lecture Demonstra¬ 

Properties, Physical 



tions 



64 

Molecular & Atomic 

Quantum Chemistry 



Structure 




Figure 14-12. Direct punch subject classification code in inorganic chemistry. 


327 



328 


PUNCHED CARDS 


the use of spectroscopists and analytical chemists. By punching the number 
79 (the atomic number of gold) in this field, it can be shown that the 
abstract is concerned with an analytical technique for gold. Other elements, 
designated by appropriate punches on the side of the card, might be used 
to indicate interfering elements. A number was assigned to each file in 
order to avoid confusion between a number of files in the same office 
using different codes but the same card. This number is punched in a single 
field at the upper left of the card. The sides of the card carry both elemental 
and numerical designations for each hole. The bottom of the card has been 
left open for entries specific to any file being developed. Space has been 
reserved for a 100-subject breakdown in the lower corner of the card with 
the remainder set up for numerical coding should such be desired. Fre¬ 
quently recurring subjects are punched in the outer holes. The deep punch 
is reserved for those which occur less frequently. Often a reference appears 
in which naturally-occurring organic substances are related to various 
chemical elements—a situation in which it is desirable to use not only the 
subjects but also the elements direct-coded on the sides of the card. Al¬ 
though an attempt to apply two codes to the same series of holes some¬ 
times leads to difficulty, it is possible to do so with a minimum of ambiguity. 
Should a reference occur for which it is necessary to code both subject 
and element on the sides of the card, it must be indicated that such an 
abnormal situation exists. Punching the hole as indicated in the lower left- 
hand corner of the illustration (Figure 14-13) shows that one or more ele¬ 
ments are coded along the sides. Although a system such as this leads to 
the isolation of a number of cards which must then be sorted by hand, it 
has the great advantage of doubling the number of entries that can be made 
on the sides of the card. The hole marked “reprint” is punched to show that 
a copy of the paper referred to on the card is already available in the au¬ 
thor’s file. Each reprint is cemented into a separate folder and numbered 
and the number is noted on the related punched card. 

In 1951 the National Association of Corrosion Engineers (NACE) 
initiated an Abstract Punch Card Service in which subscribers are pro¬ 
vided with almost 2,000 coded corrosion abstracts per year. These are 
printed on punched cards which are pre-punched for subject matter by 
the NACE. The cards are so marked that the subscriber may punch 
them to indicate the author, journal reference and original reference date. 
The NACE abstract subject filing index used for the cards is divided into 
eight main topics: general, testing, characteristic corrosion phenomena, 
corrosive environments, preventive measures, materials of construction, 
equipment and industries. A 5 x 8 inch McBee Keysort card with a double 
row of holes around the perimeter is used. The holes in the outer row along 
the top of the card are numbered 1 through 28 and notched by the NACE 



REVIEW OF APPLICATIONS 


329 



Figure 14-13. Keysort card used for the maintenance of a bibliography in geo¬ 
chemistry at the U. S. Geological Survey. 


for the abstract subject. The inner row of holes along the bottom of the 
card is for the subscriber’s use to designate journal reference and date. 
Both the outer and inner row of holes along the left-hand side of the card 
may be used to identify the authors of the reference 39 . 

Petroleum Industry 

The Petroleum History Project at Northwestern University was estab¬ 
lished in 1954 by a grant from the American Petroleum Institute to prepare 
a complete history of the industry in America and an analysis of its effect 
on American life. The study is to culminate in 1959 with the publication 
of a two-volume history. To organize and standardize its collection of 
information and make it available to all members of the project, it was 
decided to use an edge-punched card system. This system is in a dynamic 
state: cards are being used constantly and subject interests vary from 
time to time. McBee Keysort cards were used with double rows of perfora- 

w Mathay, W. L., and Hoxeng, R. B., “A Classification and Filing System foi 
Corrosion Literature,” Corrosion, 12 (11) 588-592 (Nov. 1956). 




Petroleum History Project. 


Subjects and Dates of Coverage 


Top of card (“O indicates Outer Row) 

1 Bibliography 
01 Federal 

2 Biography 

02 Inter-, intra-state, local 

3 

03 Government document 

4 

04 Patent 

5 

05 Statistics 

6 

06 Court case, law suit 

7 

07 Periodicals and serials 

8 

08 

9 Kerosine 
09 Crude oil 
10 

010 Gasoline 

11 Medical and other uses 
Oil Lubricants 
1 

012 Other petroleum products and f uels 

13 Penna., Ohio, Allegheny, New 
York, West Virginia 

013 Atlantic Coast (including Phila¬ 
delphia, Penna.) 

14 Mid-Continent, Gulf 
014 Rocky Mountains 

Left margin of card 


15 West Coast 

015 Foreign (and export) 

16 
016 

17 Exploration and drilling 
017 Production 

18 Geology, geography, history 
018 Other sciences and technology 

19 Refineries 

019 Refining and distillation 

20 
020 

21 Railroad 
021 Pipe lines 

22 Transportation (other) 

022 Tidelands 

023 Standard Oil Co. 

24 Minor companies 

024 Other major companies 

25 

025 Construction 

26 Finances, earnings 
026 Costs 

27 Securities, speculation 
027 Prices 

28 Estimates, resources 
028 Marketing 

29 

029 Supply 

Right Margin of card 


30 

31 Labor, employment 

32 Safety, accidents, health 

33 

34 Insurance 

35 

36 Public relations 

37 

38 Waste, waste disposal, pollution 

39 Conservation 

40 

41 Production rate, productivity 

42 

43 Illumination 

44 

45 Other related or competitive prod¬ 
ucts and industries 

46 


47 

48 Investigation (trust, government) 

49 Regulation 

50 Laws and legislation 

51 Competition 

52 Integration 

53 

54 Management 

55 Associations and trade agreements 

56 

57 Social impact 

58 Taxation 

59 

60 Research 

61 

62 Inspection,testing,standards,quality 

63 Equipment and packaging 


Bottom of 

card (inner row) 





1 

1921 + 

11 

1886-1890 

21 

1936-1940 

2 

... -1845 

12 

1891-1895 

22 

1941-1945 

3 

1846-1850 

13 

1896-1900 

23 

1946-1950 

4 

1851-1855 

14 

1901-1905 

24 

1951-1955 

5 

1856-1860 

15 

1906-1910 

25 

1956+ 

6 

1861-1865 

16 

1911-1915 

26 

....-1872 

7 

1866-1870 

17 

1916-1920 

27 

1873-1893 

8 

1871-1875 

18 

1921-1925 

28 

1894-1911 

9 

1876-1880 

19 

1926-1930 

29 

1912-1921 

10 

1881-1885 

20 

1931-1935 



Figure 14-14. Subject list 

used at Petroleum History Project 

at Northwestern 


University. 


330 



REVIEW OF APPLICATIONS 


331 


tions at the top and bottom and a single row at each side. Subjects are 
punched along the top and both sides of the card. The complete subject 
list is shown in Figure 14-14. Publication dates are punched into the card 
in two fields at the bottom, in the outer holes. In this system the publica¬ 
tion date is not as important as the date of coverage. For instance, a paper 
published in 1920 may contain interesting figures on the industry during 
the years 1900-1915. The date of coverage is indicated by punching the 
appropriate hole in the inner row at the bottom of the card. The author’s 
name is coded in three fields at the bottom of the card using the outer 
holes 40 . 

Since 1943 the American Petroleum Institute Research Project 44, at 
Carnegie Institute of Technology in Pittsburgh, has been concerned with 
collecting, analyzing, calculating, and compiling selected values of physical 
and thermodynamic properties and mass spectral data on hydrocarbons 
and related compounds. As of June 30,1955, the tables of the API Research 
Project 44 cover 1400 different compounds and include more than 150,000 
individual numerical entries. These are available on 45,430 IBM punched 
cards. The cards carry a two-line interpretation which permits the in¬ 
formation on each card to be read at a glance. Each compound was assigned 
a number which is given together with the full name on a name card. The 
first three columns of the name card show the group number to which the 
compound belongs. Each class of compounds is assigned a number which 
is punched into columns 9, 10 and 11. Columns 4 to 8 and 12 through 24 
are left blank on the name cards. Column 25 contains the card number. In 
most cases, only one name is given and the card number is “one.” When 
compounds have two or more names, the names are punched on separate 
cards and the cards are numbered 1, 2, etc., in column 25. The name of 
the compound is punched into the card starting with column 27. The holes 
on the data cards are assigned as follow's: 1-5, table number; 6, footnote; 
7-8, year of latest reference of data; 9, 10 and 11, compound number; 12, 
state (gas, liquid or solid); 13-23, stoichiometric formula; 24-26, card 
number. The actual property values for each compound are punched in the 
cards starting with column 27 41 . 

The Information Services Division at Ethyl Corp. in Detroit maintains 
a file of information on additives used in hydrocarbon or oxygenated- 
hydrocarbon fuels or in natural or synthetic lubricants. This informa¬ 
tion is recorded on Remington Rand punched cards and searches are 
made using the Remington Rand mechanical sorter. The file covers the 

40 Krull, A. R., “Punch Card System for the Petroleum Industry,” Petroleum Eng., 
E27-29, E32, E34 (March 1956). 

41 Sherman, J. “Physical Data on Hydrocarbons,” Petroleum Refiner, 32, (10), 
145-149 (Oct. 1953). 



332 


PUNCHED CARDS 


United States patents and Ethyl Corporation technical reports on the 
subject. The information found in patents and company reports is ab¬ 
stracted and all compounds and classes of compounds cited are listed. In 
addition, the functions of the additives and the petroleum products in 
which the additives are used are noted. The abstracts are digests or nota¬ 
tions of those parts of the reference pertinent to the file. From these ab¬ 
stracts two punched-card files have been prepared, a subject file and an 
author (or patent assignee) file. One subject card is prepared for each 
compound mentioned in a reference. The first five columns may be punched 
with an abbreviation of the name of the country, in the case of patents, or 
with the year date in the case of all other types of literature. The next 
seven columns contain either the patent number or the accession number 
assigned to the abstract. For column 13, a special code is employed for 
various types of Ethyl Corporation material, to indicate whether the 
particular abstract deals with a formal report, correspondence, or lab¬ 
oratory test data. 

Column 14 can be used as a guide to some special types of information, 
to indicate for example that the compound coded on the card is not an 
additive itself but that it is reacted with something else to produce an 
additive. The number of functional groups in the compound is punched in 
column 15. A code is punched in column 16 to indicate the elements con¬ 
tained in the compound. The area extending from columns 22 to 45 is 
devoted to the code which describes the compound cited as an additive. 
Columns 46 to 70 are used for the codes which then define the functions of 
the compound, such as antioxidant or antiknock agent. In column 80 is 
coded a notation of the type of reference in which the compound was 
found—journal articles, government material, Ethyl material, etc. Finally, 
columns 81 to 90 are punched to show the type of petroleum products to 
which the compound is added. Other areas of the card are not in use at 
present. A modification of the chemical code developed by the Chemical- 
Biological Coordination Center is used. 

One author card is prepared for each author or inventor. The company 
to which the patent is assigned or which employs the author is also punched 
into the card. Auxiliary files include a master file of all materials consulted 
in the course of gathering information for the punched-card file, including 
the notation of non-pertinent references that have been checked. The 
punched cards are filed according to a rough classification of the com¬ 
pound types, based on the atomic components of the compounds. Thus, 
compounds containing carbon, hydrogen and oxygen are coded with “H” 
in column 16 and are filed together. Cards of different colors distinguish 
the number of structural groups in the compound. The author file is 



REVIEW OF APPLICATIONS 


333 


duplicated and filed alphabetically by author and alphabetically by 
company 42 . 

The Humble Oil and Refining Company, Baytown, Texas, uses 5x8 
inch Keysort cards with a double row of perforations for a catalyst file. 
This file was adapted for use in a technical man’s personal file for in¬ 
formation on a wide variety of subject matter. There are 56 numbered 
holes available for indexing at the top and bottom of the card as well as 
letters and symbols on either side for more specific coding. A coding system 
using two code numbers per subject was selected. The numbers are assigned 
to the subject by means of a random number table. This allows therefore 
a total of 1,540 separate items to be indexed in the file and as many as ten 
items to be indexed satisfactorily on a given card. In 1954 this file was 
indexed for 150 major subject headings and contained about 600 cards 
which represented about 3,000 separate items. Three fields on each card 
are used to code the authors name, the first two letters of which are punched 
directly into the alphabetical index. The top and bottom of the card are 
devoted to the subject index and the left side to a formula index 4 *. 

Hobbies 

McBee 5x8 inch edge-punched cards have been used to great advantage 
by T. T. Hill of the Edwal Laboratories in Ringwood, Illinois, to prepare 
topical exhibits of stamp collections and to index any technical data about 
them of interest to the specialist. From one to ten stamps can be mounted 
on the face of a card with the name of the country and the date typed on 
the card. Technical and descriptive information is coded and the file can 
be hand-sorted for information on such things as perforation method, type 
of ink and paper, subject, historical connection and type of stamp 44 . 

A noteworthy example of the variety of uses made of hand-sorted punched 
card systems is an application to contract bridge. The Bridge Hand of the 
Month Club, Inc., supplies two decks of playing cards perforated and 
notched along the narrow edges with three holes in each position and ten 
positions along each end. Needle sorting each hole in one position separates 
the four predetermined hands. After a set of four hands is sorted, the hands 
are played and scored in the usual way. The result is then compared with 

41 Graham, M. H.; Hildenbrand, B. S.; and Weil, B. H., “Indexing of Fuel and 
Lubricant Additives By Machine-sorted Punched Cards,” presented before the Ameri¬ 
can Chemical Society, Division of Chemical Literature, Dallas, Texas. April 11, 1956. 

41 Hoffmann, E. J., “Use of Punched Cards For Filing Technical Data,” Humble 
Oil & Refining Co., Texas Chapter Bulletin, 5, (4) 10-16, (May 1954). 

44 Hill, Thomas T., "Stamp Collecting and Punched Cards,” Private communica¬ 
tion to J. W. Perry, August 11, 1953. 



334 


PUNCHED CARDS 


the recommended bidding and play discussed in the accompanying in¬ 
struction book 45 . 

It has been suggested that punched cards are also admirably suited 
to filing and retrieval of interesting and instructive hands. Cards held 
in a given situation, bidding and play, plus comments and analyses, 
can be clipped from magazines or newspapers and pasted on the cards. 
These in turn can be coded according to the players or to the various spe¬ 
cial features of the bidding and play. 

A suggestion from the Edwal Laboratories involves the use of punched 
cards for indexing photographic slides. The holes used for coding are 
punched directly along the edges of the slide itself. Ten classifications are 
possible with five holes punched at the top and five at the bottom of the 
slide. By turning the box upside down and removing the bottom instead 
of the top, the lower row of holes may be needled in the conventional 
way. The file may be searched for subject, date, locality, etc. A master 
code card indicates the meaning represented by each notch. For example, 
a code card which reads, “Glacier Park, 1940 T 3 ” shows that all pictures 
taken in that locality at that time are notched in the top of the slide, third 
hole from the left 46 . 

Another system for indexing photographic slides involves numbering 
each slide as it comes in from the processor and giving the same number to 
a corresponding punched card. A code of numbers is assigned to an alpha¬ 
betically arranged list of subjects and these numbers are notched on the 
card as they apply to the slide it represents. For example, # 1 on the sub¬ 
ject list might be architecture and any slide whose subject is architecture 
will be notched in the first hole on the card. Up to 19,999 subjects can be 
recorded on the upper as well as the lower edge of the card. The ends can 
be used to code the year in which the slide was made. On the face of the 
card is typed the location of the subject, the shutter settings and the date 
the picture was taken. This system uses 5x8 inch McBee cards with a 
single row of perforations along the edges 47 . 

Miscellaneous 

Included here are some of the many interesting uses of punched cards 
which, because of their subject matter or approach, do not fit into the pre¬ 
ceding groupings. 

u This system has been devised by the Bridge Hand of the Month Club, Inc., 
28-36 214th Place, Bayside, New York. 

44 Patton, A. R., “Punch Card Filing System for Your Slides,” The Camera, 73 (1) 
63, 130, (Jan. 1950). 

47 Davis, L. R., “Locate Your Slides and Negatives With This Punch Card File 
System,” U. S. Camera, 16, (9) 68-69, (Sept. 1953). 



REVIEW OF APPLICATIONS 


335 


Arthur D. Little, Inc. of Cambridge, Mass., has found two new applica¬ 
tions for punched cards. One is a coded and punched file for all data collected 
on explosives from 1950 to 1955. The literature group of the company has 
also established a McBee card system to classify company personnel by 
educational background and experience, thus enabling more efficient use 
of their personnel on various research problems 48 . 

The New York Society of Electron Microscopists has issued a bibliog¬ 
raphy on Keysort cards which will keep abreast of the literature in all 
fields of electron microscopy. An outstanding feature of this bibliography 
is its ease of use. It is already coded and punched by the bibliographer. 
Articles containing multiple subjects are easily found by each subject as 
well as by author. The first issue covering the years 1950-52 (approxi¬ 
mately 700 cards) was available at the time of the publication of this 
notice 49 . Early publication of material for 1953 and quarterly publications 
on current literature thereafter are planned. 

The National Intern Matching Program acts as a central clearing agency 
for hospitals seeking interns and students seeking internships. Each student 
submits a list ranking hospitals by preference. This information (about a 
30,000 item cross-index) plus quotas for each hospital, forms the input 
data for the IBM 704 which performs the actual matching. The result of 
the 704 operation is a matching of preferences of the hospitals and students 
so that each student gets the hospital of his choice. This is determined by 
the way the hospital ranks him and its quota. The 704 thus analyzes 30,000 
applications to approximately 800 hospitals which have been named by 
7,000 students. The actual 704 running time of 1956 matching was 1 hour 
and 45 minutes 60 . 

The Massachusetts Institute of Technology has compiled a bibliog¬ 
raphy of all important world literature on coffee for the Coffee Brew¬ 
ing Institute, Inc. of New York using specially-printed edge-punched 
cards (Figure 14-15). E-Z hand-sorted cards are used to index un¬ 
coded information on the date of publication, source of reference, sub¬ 
ject, author, title, publisher and abstract of the reference. Although the 
cards do not have to be kept in any specific order for sorting, they are given 
a number so that they can be arranged in order when a list is prepared for 
distribution. The year of publication is set up as a simple code which makes 
it possible to select 399 separate years. A hole marked “completed” in the 
upper right-hand corner is punched only after a final check has been made 

4g 4 ‘Current Research and Development in Scientific Documentation / 9 compiled 
July 1957 by the Office of Scientific Information, National Science Foundation. 

49 “Bibliography on Electron Microscopy/ 1 Science , 118, (3066), 378, (1953). 

M Personal correspondence from Joan R. MeJoynt, National Intern Matching 
Program, Inc., Chicago, Illinois, to B. L. Haksteen, July 8, 1957. 



336 


PUNCHED CARDS 



Tl*« Of fUOllCATIOM 

CO fftl LITCMTU*r*UlfVEY~ 

COFFEE SKEWING INSTITUTE. INC. 

FOOO TECHNOLOGY DEPT. - MAT. 


• : a • L • 

PI “ | 


"•P«P 


ijiMM i liji x[i 

»r»i«r»TTTTt« 


Idfili 


WI1MT WtJtCT WOH 


III M[l» 1 

ItWV 


r 

\t 5 

Eg 

•JJ- 



h mi 

UMKtl 


i, 



Black, J. W. 

Freshly ground coffee and "blown" tins. The Analyst 51: 403-404. 1926. 

A quantity of Costa Rica coffee was ground after keeping for 8 days froa 
the roasting tlae, and the evolution of gas laaedlately determined. For 200 g. 
of coffee, 52 cc. of gas was collected In 1 hr.. 90 cc. In 5 hrs., and 132 cc. 

In 48 hrs., and this result Is regarded as typical. 

—Br. Ch. Abr. 

The evolution of gas froa ground coffee Is probably not due to the action 
of air on the coffee but is occasioned by the gradual ellalnatlon of gas froa 
the coffee, which was evolved during the roasting process but held under pressur^ 
In the roasted bean. The aat. of gas evolved varies with the degree of grinding 
the severity of roasting and the lapse of tlae. 

—Ch. Abr. 

Abstract In Zeitschrift fur Lebensaittel-Untersuchung und Forschung 61-62: 
541. 1931. 


’4 

r . ^ 

«••• 
•• • 


Figure 14-15. Card used to index information on the world literature on coffee. 


and the information punched is confirmed. Some 10,000 references were 
accumulated in two years covering the period 1925 to 1949, and for each 
individual year from 1950 through 1955. Work continues on the 1956 ref¬ 
erences and for the years prior to 1925 M . 

Daily copies of all crime reports in the Los Angeles Police Department 
are posted to various daily charts which are used to prepare periodic statis¬ 
tical reports. The crime reports are then coded according to division of 
records number, date of occurrence, date reported, location, type of crime, 
who, what and where attacked, means and object of attack, trademarks, 
and description. The coded items are then punched into two types of IBM 
cards; property loss cards and miscellaneous complaint cards (Figure 14-16). 
A card is prepared for each separate offense. The cards are used for the 
preparation of routine and special reports, and have been an invaluable aid 
in the analysis of modus operandi to identify suspects and to locate possible 
suspects already in custody for some other offense 52 . 

Specially designed 8)2 x H inch McBee Keysort cards (Figure 14-17) 
have been printed for preliminary studies of the occurrence and character 
of deep-sea diving accidents by the Navy. This card was selected because 
of the limited number of the eventual total sample—under 10 , 000 . The 

51 Lockhart, E. E., “A Card Punch Bibliography on Coffee,” report submitted to 
J. W. Perry, May 14, 1956. 

,J “Modus Operandi as Developed by the Los Angeles Police Department,” a 
report prepared by the Statistics Unit Planning and Research Division, LAPD, 
December 1955. 























REVIEW OF APPLICATIONS 


337 


22 

313 31 


*«J 

mkiSi 


•p •{• 

i 


V 

k2|2 2| 


* 




n 

C 4 4 4 
555 


IHHIIM 


2 2 2 | 
31 

4 4 

5 S 


1 

2 2|2 
323 
444 
S|5S3 


i l 7 7|7 7| 7 

• I 


ill 


l|l I 1 *1 lit 

*|2 2 2 J 2 2p| 

2(3 3 3*3 3| 

^ssHs) 

1 1 M *1 

l| 7 7 7»? 7[?| 

■•Ml 


WMiifll 


22 

33 

44 
NS I 
II 
7 7 

[III 11 


22 2 


3 3 ill 3 93 9 3|J 3|3 3 3 


2 2 212 2 212 2 


I I I I I I 1 I I I lh 1 I I I I I 11ll 


sssssisssl 

lllliqi tt|ll| 

7 7 7 

[HI 


ll|lll| 


|l I li I Ml 11|0 III I Ml I 111 IM l| 


2 2 2b 2 2 22 2 2 »2 2 32 2 2 


i ill 


inbnlmDM 


4 4 4 4 4 4|4 4 4| 


«V»I| 99 999 999fMltf»9 VVi « 9 Ml 1 ill I19 MS I »ll 9 9'1 1911 91Mil 99199 V% 9*9119 9 f 1 1 9 9 9 9W99! 


|7 7 7|7 7 7 7 1 7 7 7 7D 

in nil 


SSSkSShM 


mi 


mavi ■■ ub mm • a v a ■(• n H n a 

11111111 in 1 


1 3 3 3 3 3 3 m 
4444444444 
5535535955 

IllllilSIl 


7 7 7J7 7 

Hill 


if 


1 m 1 ih 1 m I 
2 2|2 2|2 2 2 2 2 2|2 2 2 2 2 2[2 2 2 2 2 
3 3 3 3 3 3|3 3 3 3 3 3|1 3 3 3 3 




7 7 7 7 7 7 

nun 


iki 


|4 4 4 4 4 4|4 4 4 4 4 4|4 4 4 4 4 

sstssslisiissssssi 

mmiiiM 


I 


/T 

to’l 

it 1 

ll 

! 

Ui 

1 

h 

1 

i| 1 |j 

k 

ill 

I 

1 

« 

Sm 

Ij* 

Sal 


1 

£ 

• HIM 

111« • • 
111111 

222222 

| 333333 

1 4 41444 

9 19 9 S 5 

• Hill 

171777 

• •Hill 
91 • 11 ill 

i • 

If 
• • 

It 

in 

1 1 

12 

13 

14 

55 

II 

7 7 

II 

991 

• MB 

bi 

a m 

1 1 

12 

13 

1 1 4 

135 

ill 

77 

II 

99 

r*i we 

II 

III! 

1 1 

72 

13 

14 

15 

IS 

17 7 

III 

99 

Itu 

• 

14 

• Hi 

inti 

■ ••bn 

I Mil 1 

f 2 2*2 2 
• 

nia 3 

1 4 4*4 4 

l 

ISStSS 

Ill'll 

1 

7 7 7i7 7 

Ill'll 

i 

99*11 

111 

»bn 

II 1 

iVl 

i 

It3 3 

i'll 

i 

5d5 

i'll 

i 

7|7 7 

III 

« 

Mil 

nan 

II 

ma 

1 1 

722 

133 

144 

355 

III 

II 

99 
• *• 

HID 

1 1 II 

7 2 2 7 

3 3 3 3 

1444 

5 5 5 5 

III! 

7717 

till 

9911 

p saa 

III 

a a >> 

1 I 1 

727 

333 

4 4 4 

155 

III 

1 7 1 

1 II 

Ml 

na 

9 D I 
ih 

1 1 1 

22 

333 

444 

555 

SI 

71 

1 

9 

in 

• • 

MB 

1 1 

ID 

«i« 

727 

333 

444 

535 

SIS 

777 

HI 

991 

II 

• «< 

1 1 

72 

333 

444 

535 

SSS 

777 

SIS 

999 
• »» • 

II 

iwm 

1 1 1 

7 22 

333 

555 

SIS 

77 7 

III 

9 9 9 9 
am 

ib! 

m 

7 77 

3 33 

144 

555 

ISS 

777 

III 

ill 9 

m 

1 

ma 

1 1 

722 

333 

144 

555 

SIS 

7 77 

991 

aww 

lllll•llllllll••llllll ^ 

II 1 1 11 1 1 1 1 1 1 1111111111 | 
2227722222772222222277 & 

3333333333333333333333 g 

4 4 4 4 4 4 4 4 4 4 44 4 4 4-4 4 4 4 4 4 4 | 
5555555555535555555553 | 
SSSSISIIIIIIISSSIIIIIS I 
7777777777777777777777 q 
llllllllllllllllllltll Jj 
9999999999999*99999919 J 


Figure 14-16. Punched cards used to record crime reports in the Los Angele 
Police Department. 

card has space for writing in the identity of the patient and for special 
notes 5 *. 

Group feeding operations are increasing as is indicated by research sur¬ 
veys conducted in the field of agricultural economics. Nutritional research 
is also advancing, and the application of this knowledge and the develop¬ 
ment of a technique by which food nutrients may be rapidly and accurately 
calculated, may lead to the adaptation of manually operated marginal 
punched cards 54 . 

Dr. Mary K. Bloetjes, Professor of the Department of Institutional 
Management, New York State College of Home Economics, uses punched 
cards as a teaching aid in her course on Cost and Production Control. 

** Personal communication from H. W. Gillen, Physiology Branch, U. S. Naval 
Medical and Research Laboratory, New London, Conn, to J. W. Perry, March 27, 
1956. 

M Bloetjes, Mary K., “Management Research in Food Service Operations,” 
presented at the 37th Annual Meeting of the American Dietetic Association in Phila¬ 
delphia, October 29, 1954. 



























































































338 


PUNCHED CARDS 



Figure 14-17. Punched card designed for Navy to record deep-sea diving accidents. 


This course is taught by the case problem method, using a series of eight 
menu items analyzed for the various factors indicated on the card such as 
form of purchase, type of food, condition of food, amount of or absence of 
waste, etc. The cards are direct-coded in order to facilitate teaching 55 . 

The Bureau of Aeronautics of the Navy has a service test under way for 
the ultimate elimination of blueprints of engineering drawings and the 
substitution of microfilm mounted in electric accounting machine (EAM) 
cards. The EAM cards are punched and interpreted for each exposure of 
microfilm, and will contain the drawing number, microfilm frame number, 
Federal supply code and model designation of equipment. This information 
will be repunched into Filmsort aperture cards which will be handled by 
standard punchcd-card procedures 56 . 

FACSI Incorporated, Deerfield, Illinois, has developed a unique refer¬ 
ence system. FACSI is a code word that represents the group of words 
Fast Access (of) Coded Small Images. The system combines edge-punched 
cards, a code specially designed for nondestructive testing literature, and 

65 Personal communication to R. S. Casey from Dr. Mary Bloetjes, November 4, 
1955 and November 16, 1955. 

46 “Military Specification. Microfilming Engineering Drawings and Related Data, 
Requirements For,” MIL-M-18872 (Aer), June 13, 1955. 






















REVIEW OF APPLICATIONS 


339 


articles printed directly on the cards. Each film size (8 x 10}% inch) card 
contains a perfectly readable (reduced 6 to 1) image of a complete NDT 
article. All cards are pre-punched with the proper code and may be kept 
in random order. Every article published in the Journal of the Society for 
Nondestructive Testing has now been reproduced and stored on a specific 
card. The notched hole coding permits the extraction of any card in a few 
moments 67 . 

Since this chapter does not attempt to be exhaustive but merely indica¬ 
tive of the varied uses of punched cards, there were many interesting arti¬ 
cles which came to our attention that have not been discussed in the pre¬ 
ceding sections. These are listed in footnotes 58-69. 

47 Staats, H. N., “Data Extraction in Nondestructive Testing,” Nondestructive 
Testing (Jan.-Feb. 1957). 

48 Lenihan, J. M. A., “Isotope Catalogue on Punched Cards (Edge-notched),” 
Brit . J. Appl. Phys. 3 (29) (1952) ( Nuclear Data.) 

49 Way, K. “Data type Abstracts,” Physics Today , 10, 17-18, (1957) ( Nuclear Data.) 

90 Schwabe, C. W. and Davis, L. R., “Marginal Punched Cards in Veterinary 

Research,” Am. J. Vet. Research 15, (57) 634-638, (Oct. 1954) (Biological Data). 

81 Gey, K. F.; Kalbe, H.; Schon, H.; and Stegemann, H. “Documentation of 
Physiological-Chemical Literature on Punched Cards.” Hoppe-Seyler's Z. physiol. 
Chem. t 301, 70-77 (1955) (Biological Data.) 

69 Reumuth, H., “The Indexing of Chemical Compounds. A Contribution to the 
Problem of Organization of the Literature,” Z. Angew. Chem. y 41, 1204-7 (1928). 
(Laboratory Records.) 

63 Preliminary Report on Research in Progress in Scientific Documentation, 
compiled August 1956 by the Office of Scientific Information, National Science 
Foundation. Section on Monsanto Chemical Company. {Laboratory Records.) 

84 Kountz, R. R., “IBM Punch Card Data-control in Pilot Plant Operation,” 
presented at the American Chemical Society Division of Water, Sewage and Sanita¬ 
tion, Fall 1952. {Laboratory Records.) 

84 Jones, W. S. and Butterfield, P. H., “A Technical Information Service Using 
Punched Cards for Indexing and Retrieval, ,, presented at the American Chemical 
Society meeting in Minnesota, September 12, 1955. {Technical Services.) 

88 “Guide to NACE Corrosion Abstract Punch Card System With Appendix A, 
Sections I-VI,” published by the National Association of Corrosion Engineers, Pub¬ 
lication No. 51-6, June, 1951. {Chemical Literature.) 

87 Demer, L. J., “Bibliography of the Material Damping Field.” WADC Technical 
Report 56-180, June 1956, Wright Air Development Center. {Chemical Literature.) 

88 Peakes, G. L., “The Unit Card System in the Indexing of Internal Technical 
Reports/’ Chapter 11, pp. 149-164 in “Progress Report in Chemical Literature Re¬ 
trieval,” edited by G. L. Peakes, A. Kent and J. W. Perry, New York, Interscience 
Publishers, Inc., 1957. 

89 Peakes, G. L., “Experience with the Unit Card System for Report Indexing,” 
Chapter 19, pp. 306-327 in “Information Systems in Documentation,” edited by J. 
H. Shera, A. Kent, and J. W. Perry, New York, Interscience Publishers, Inc., 1957. 



Chapter 15 


A CASE HISTORY OF A ZATOCODING 
INFORMATION RETRIEVAL SYSTEM 


Claude W. Brenner 
Allied Research Associates, Inc., Boston, Mass. 

AND 

Calvin N. Mooers 

Zator Company, Cambridge, Mass. 


The Problem 

A rapidly growing collection of research reports presented an acute 
reference problem to Allied Research Associates, Inc., Boston, Massachu¬ 
setts, in 1954. This organization of engineers and scientists, doing research, 
engineering, and development in the aeronautical and physical sciences, 
had been expanding since its beginning in 1951, with three engineers on 
its staff. At first, personal files of reports were sufficient for the company’s 
information filing and retrieval needs. Later, project files were set up, and 
reports touching on the different projects were segregated into these files. 
As more contracts were undertaken, the engineering staff increased, and 
the rate of influx of technical reports and papers steadily mounted. By late 
1954 the company had a staff of fifty, and the bulging files held about 3000 
reports, with more coming in every day. 

It was evident that the files would very soon become unmanageable. 
The first step was to turn to conventional library techniques, and a library 
school graduate was hired. It was to be her job to organize the company 
report collection so that it could be used easily by the engineers. It was 
hoped that she would be able to merge the catalog cards that had been 
received from ASTIA, AEC, and NACA. This required considerable 
knowledge of the subject matter and it was soon evident that in order to 
do the job extensive assistance would be required from the engineers. Not 
all of the incoming reports had catalog cards, and the analysis of their 
contents could not be left completely to the librarian. Moreover, it was 
the impression at the company that even had it been possible to interfile 
the various cards, the burden of filing multiple cards for each report would 
soon have become intolerable. The company therefore decided that a card 
catalog system would not be adequate to serve its needs. 


340 



ZATOCODING INFORMATION RETRIEVAL SYSTEM 


341 


Preliminaries to Setting up a Zatocoding System 

At this point the engineer who supervised the library operation learned 
of the Zatocoding system. He and several other engineers witnessed a 
demonstration and tried sorting a pack of Zatocards. They also did some 
additional investigating. They contacted clients of the Zator Company to 
see how their systems had worked out. They studied the available books 
on punched cards. They got prices of other cards and equipment. They 
checked to see what assistance salesmen of other equipment could give in 
setting up a retrieval system for Allied Research’s highly technical field. 
They estimated the likelihood that a conventional library system would 
be satisfactory. Probably the most decisive reason behind their final 
choice was the technical guidance provided by the Zator Company during 
installation of the system (Chapter 3). 

The Zatocoding System 

The Zatocoding system has three parts. There is the strictly mechanical 
part represented by the Zator “800” Selector and by the edge-notched 
Zatocards. This is the most tangible part of the system, though in some 
ways it is the least important. The second part is the technique of using 
random-like descriptor code patterns and of notching these code patterns 
into the edge of the card in superimposition. This is the Zatocoding tech¬ 
nique. The third part is by far the most important and requires the most 
explanation. It is the system of “descriptors” by which documents are 
characterized, and by means of which retrieval questions are turned into 
prescriptions for search. 

One card is made up for each of the reports in the collection. Notches 
along the edges of the cards permit a mechanical sorter to scan the cards 
and to select some of them. The subject content of each report is related 
to the pattern of notches in the card by the coding scheme. Therefore the 
sorter is able, by a strictly mechanical process, to select cards from a pack 
according to subject matter. All the cards are scanned for each retrieval 
question. Complete scanning has the advantage that the cards need not 
be kept in any order. Card filing is thus eliminated. 

The Zator “800” Selector 

Figure 15-1 shows the Zator “800” Selector 1 in operation. A pack of 
about 200 cards is placed in the* black, box-like upper part of the selector. 
The box is vibrated by a small motor, as shown in Figure 15-2. Near the 
bottom of the box are rods or needles, which run from front to back. Each of 
the rings shown in the Figure is attached to a rod. It is easy to pull the 

1 Mooers, Calvin N., and Charlotte Davis Mooers, “Card Selecting Device,” U.S. 
Patent No. 2,665,694 (1954). 



342 


PUNCHED CARDS 



Figure 15-1. Sorting cards with the Zator “800” Selector. Cards are taken from 
one of the side trays, arc sorted, and then are placed in the other side tray. The ac¬ 
cepted cards are dropped to the table in front of the machine. 

rods out by means of the rings and to insert them again in a different 
selective pattern. 

The Zatocards, like the one shown in Figure 15-4, have notches along 
the edges representing different subjects. In making a selection the pack 
is placed in the selector machine with one of the notched edges resting on 
the sorting rods. Most of the cards in the pack rest on the top of the grid 
formed by the rods. However, some of the cards, as shown by Figure 15-3, 
have notches in the position of each of the selector rods and are not sup¬ 
ported on top of the grid. These shake down a little way from the rest of 
the pack, and are the desired cards. 


ZATOCODING INFORMATION RETRIEVAL SYSTEM 


343 



Figure 15-2. Cross-section of the Zator “800” Selector showing the vibrating 
motor and the manner fn which most of the cards stay on top of the selecting rods. 



Figure 15-3. Diagram showing how the cards whose notches fit the pattern of the 
selector rods drop from the rest of the pack. 

Looking again at Figure 15-3, it is seen that the pack of rejected cards 
can be engaged by a rod or tool inserted through the holes near the top 
edge of the cards. The desired cards, having dropped down a little, are not 
so engaged. Thus, when the tool is raised, the pack of rejected cards is 
held on it and lifted out of the selector. The desired cards are not engaged 
by the tool and drop free from the pack to the table top. For cards with 
coding on both the top and bottom edges, the selection operation for the 
codes on the second edge is carried out after the cards have been sorted 
according to the first edge. 

About one second of the vibrating action of the selector is sufficient for 
complete separation. The speed attained in sorting depends on how nimble 
the operator is with his hands. Speeds of better than 800 cards per minute 
are easily attained. 













344 


PUNCHED CARDS 


Zatocards come in two styles. One style has notches only in a single edge. 
It has 40 notching sites. The card used at Allied Research has two edges 
given over to notches. This style has a total of 72 notching sites, and nearly 
twice as many descriptors can be notched into the double-edge cards. To 
sort the double-edge cards, the selector is first set up to scan the top edge 
of the cards. All the cards in the collection are run through the selector, 
which gives a partially selected pack of only a few hundred cards. The selec¬ 
tor is then set up for the patterns on the bottom edge of the cards and the 
small pack of partially selected cards is run through. The second sorting 
goes very rapidly because there are at most only a few hundred cards 
involved. The cards that emerge from the second selection are the desired 
ones. Most of the selection time is taken up by the sorting on the first edge. 
For this reason, the speed of sorting is almost the same regardless of whether 
single-edge or double-edge cards are used. 

Random Superimposed Codes 

The second part of the Zatocoding system is the random superimposed 
coding method called Zatocoding 2 ' 3> 4 . A pattern of notches is established 
for each subject covered by the documents in the file. The subjects to be 
coded are all overlapped or superimposed in an undivided area of the card. 
If two edges of the card are used, they are used as if they were one long 
edge. One might think that the superimposing would lead to an awful 
mix-up, but it doesn’t, provided the code patterns do not resemble each 
other too closely. One way is to use random patterns; these are patterns 
generated by flipping a coin, or by some other similar means. The Zato¬ 
coding method teaches that any patterns that are “random-like” in the 
sense that the individual code marks are well scattered and fall with 
approximately equal incidence on all the coding sites can be used for 
coding. A list of random-like patterns has been prepared for Zatocoding 
systems to eliminate the necessity of deriving new patterns for each instal¬ 
lation. 

The Zatocoding method of using superimposed random-like code pat¬ 
terns is illustrated in Figure 15-4. The Zatocodes for the various descriptors 
have been written in on the card. The first tw r o numbers represent notches 

2 Great Britain, Patented, No. 681,902, 3 September 1948; Canada, Patented, 
1956, No. 534,926, 25 December 1956; U. S. Patent pending. 

* Mooers, C. N., “Zatocoding Applied to Mechanical Organization of Knowledge,” 
Am. Doc., 2, 20-32 (1951). 

* Mooers, C. N., “Choice and Coding in Information Retrieval Systems,” Trans¬ 
actions of the Inst, of Radio Engineers Professional Group on Information Theory 
PGIT-4, pp. 112-118 (September 1954). 



ZATOCODING INFORMATION RETRIEVAL SYSTEM 


345 


heat transfer 
theoretical study 
supersonic 



high 

1302/14.24 

AS 2624 

temperature 

11.14/39.2 

University of Colorado 

heat transfer 

23,31/8.30 

Determination of the temperature distribution 

theoret. study 

?.a/4.32 

in a solid diamond-shaped wing in supersonic 

wing 

6.24/7,30 

flight, considering bcamelse heat flow only. 

cross section 

10.15/U.30 

with unsynmetrical conditions at the boundaries* 

supersonic 

3.26/4.10 

hr Broil 

fluid flow 

7.40/11.31 

UnclMtlflod 

© 




iMMir&irAllMIIIBM MBBlir aYA lBIIIIB IB 

1 f ■ " ‘ -r . jl EH ' IBBDT EXX I . I 

, /|I9 !®mlvMI.* '* ^ ^ • AIM 


transfer 


tioal study 
supersonic 

Figure 15-4. A Zatocard from the collection of the Allied Research Associates, 
Inc. showing descriptors and their Zatocodes at left and the report file number at 
upper right. The lines and arrows illustrate how the three descriptors are coded. 


in the top of the card; the second two, notches in the bottom. In the actual 
system, however, there is no need to write in the code numbers. Figure 
15-4 also illustrates the manner in which selection is performed by Zato¬ 
coding. In the case shown, three subjects simultaneously define the desired 
selection. They are “heat transfer,” “theoretical study,” and “supersonic.” 
The individual codes, and the way they are superimposed to form the total 
selective pattern, are shown by the diagram. The arrows correspond to 
the total selective pattern of rods set up in the card selector. Evidently 
selection will be made by this selective prescription, since there are notches 
in the card in every position where there is a selector rod. Note that a 
selected card may have more subjects (and thus more notches) than the 
selecting prescription. Selection extracts the cards that have at least all 
of the prescribed descriptors. 

In addition to the handful of cards that contain the descriptors that 







346 


PUNCHED CARDS 


were prescribed, there will often be two or three cards on which the de¬ 
scriptors do not correspond in any way with the prescribed ones. These 
are the “extra cards” of Zatocoding. They are harmless, because they are 
so few in number and are so easy to discard. All the cards with the pre¬ 
scribed combination of descriptors are sorted out, so no cards are ever 
missing. The fraction of extra cards is mathematically predictable, and 
can be set to as small or large a value as may be desired by varying the 
number of notches per pattern in the Zatocodes. 

The Descriptor Dictionary System 

In contrast to the rather straight-forward mechanics of the code scheme 
and the selector is the third part of the Zatocoding system. This is the 
intellectual part and is called the descriptor dictionary system; it is the 
most important part of the system. It is called a descriptor dictionary sys¬ 
tem because it is not merely a list of subject words. Instead, it comprises 
several different kinds of lists, each having a definite function. It is the 
intellectual tool that couples the mind of the information searcher to the 
hardware of the Zatocoding system in such a manner that the hardware 
does the work of selecting the desired subject matter from the file. 

Table 15-1. A Portion of the Alphabetically Arranged Scope Notes. De¬ 
scriptors are Preceded by Asterisks. Terms not Descriptors [n.d.l are Gross- 
Referenced to Descriptors. 

* Stability 135 12-11:3^-35 

In aeronautical engineering, pertains to the study of aircraft stability as used in 
conjunction with ‘Static, ‘Dynamic, ‘Lateral, ‘Longitudinal. Also refers to in¬ 
stability, such as buckling or other structural instabilities. Use with ‘Derivatives 
in stability and control studies. For Lateral-longitudinal Stability Coupling (n.d.), 
use ‘Stability plus ‘Lateral plus ‘Longitudinal plus ‘Interference. 

* Stall and Buffet 136 38-16 : 26-7 

Stall pertains to the condition of partially or wholly separated flow on air-foils at 
high angles of attack. Buffet is the disturbance due to periodic boundary layer 
separation on a surface or the motion of a surface in a fluctuating wake. 

* Static 137 38-3: 28-25 

With ‘Stability, pertains to static stability studies. 

Statistical Mechanics (n.d.) 

Use ‘Thermodynamics 
Statistics (n.d.) 

Use ‘Probability. 

Stick Force (n.d.) 

Use ‘Control plus ‘Biology. 

Strain Gage (n.d.) 

Use ‘Stress and Strain plus ‘Instrumentation. 

*S/ress and Strain 138 15-3 : 33^8 

Any process involving the loading and deflection of structures, e.g., bending of 
beams, deflections of plates, theoretical elasticity studies, elastic behavior. Use 
also for Torsion (n.d.). With ‘Instrumentation it means Strain Gage (n.d.). 



ZATOCODING INFORMATION RETRIEVAL SYSTEM 


347 


A descriptor is something like a “subject heading” in library practice, 
though it is usually much broader in meaning. For instance, a subject head¬ 
ing might be “oils—effect of temperature on viscosity.” In descriptor 
analysis, the separate descriptors “oil,” “thermal,” and “viscosity” would 
be used together to delineate this meaning. Each descriptor is a word- 
symbol standing for an idea or concept, generally of a rather broad scope. 
The particular scope of meaning for a descriptor is assigned in such a way 
that the descriptor will be most useful for retrieving information in a 
specified collection. Thus, the assignment of meanings at Allied Research 
is in part quite different from those assigned in other Zatocoding systems. 
Retrieval meanings need not conform strictly to standard technological 
usage of the word chosen to be the descriptor symbol. Because the mean¬ 
ings are often slightly different from the ordinary usage, it is essential that 
the descriptor dictionary system include a list of “scope notes,” with a 
scope note for each descriptor. An alphabetically arranged list of scope 
notes such as shown in Table 15-1 then makes the full range of assigned 
meanings easily accessible to anyone desiring to use the Zatocoding system. 
These special descriptor meanings are private, for use in retrieval only, and 
there is no intent (nor likelihood) of imposing them upon ordinary speech 
or technical writing within or outside the company. 

Deriving the Schedule of Descriptors 

The schedule of descriptors is the most important component of the dic¬ 
tionary system. At Allied Research, a panel of four top engineers and physi¬ 
cists worked together in deriving a schedule of descriptors. With this panel 
of top personnel, problems of scientific and company policy as they affected 
the future use of the retrieval system could be settled on the spot. Thus, in 
areas where the company expected to embark on a new line of endeavor, 
the group was anxious to make sure that appropriate descriptors were ob¬ 
tained. 

Deriving the descriptors is a strictly empirical process. On four separate 
occasions the Allied Research panel met with the Zator representative. A 
stack of reports, giving a typical sample of their file, was brought out and 
placed on top of the conference table. The top report was taken, its title 
and abstract were read to the group, and it was passed around for a brief 
examination of its contents. Then the question was posed, “Why would 
anyone at Allied Research be interested in using this report?” The answer 
may have been that it was about ‘propellers, that it was about propeller 
aerodynamics, and that it was a wind tunnel study. Each of these was taken 
as a presumptive descriptor and written down. The same empirical process 
was followed with the next report, and so on. On more than one occasion, 
sad experience has given convincing proof that descriptors “dreamed up” 
in an armchair without reference to actual reports are worthless. This 



348 


PUNCHED CARDS 


93 

© $ 


t * 

S 8 •- 

ft S ..2 


a S -p © *-s In 

■s 2 :§ s 111 s 

5»$-S=.§5 8s 

•« h Js '3 £ *2 ® 3, 

CQ pH Ph P JS *<! 


£ 


a s 

.2 § 

*■5 © 

o3 - 
© h 

• N V 

*& ? 

CL 55 

o3 « . 

•m ^ S 
O © » 


fl 

M C -p g ® ^ 

£ f* § 2 © a r 
jQSophO © 


T3 

8 

© 

si 


oo 

V 

o 

£ 

n. 



a> ® 
A A 


£ HHOKO 

8 
•e» 

s 
© 
si 

"© 

© 
k 
© 

© 
k. 


k 

1* 

~ o 

_ >>•£ 

JSK 

.2 s a ? s 


« *> 


© 

D 

O* 

• pH 

fl 

to 

© 

H 


© 
© 
to g 
a «? 


* o 


U 3 

.03-2 
« <5 fe 


© 
to 

3 © 

43^ S J >, 

8 *2 g ^ o 

65 S^l -S 

ilia's 2J*g 

QcQMWHflSfri 


isi 


e 5 


*0 

> 


r© 

© 


si 

© 

© 

*© 

S3 

© 


A 

to 

© 


*© 

2 
00 

© •* to 


©• 

£ 

C 


© 

to 


>» 

2 o 

-5 ki © 


© 

£ 


*r» 

00 


03 


T3 

•a .g u § 

a 5 *a "8 -2 -g 

I •§>*•£ || 

^ o © © © .© 


■8 g 
mOm 


'S 

2 

©» 

^ * 


8 | 

"I 


2 

3 


JS _ .2 .§ © -2 

j ® •» i 6 § 

« 3 Ss "3 s 

S O S ■< S H <! 


03 

a 

?S> 


© 


•2 


* 

o 

£ to 

„ <S ”3 
9-0 S 

E*9 


© .2 
.2 *3 a 

a § 2 

O m £ 

i &6'L 

*" aiSs 
HtcB JH 



8hock waves 
Finite span 



ZATOCODING INFORMATION RETRIEVAL SYSTEM 


349 


empirical procedure of discovering descriptors is surprisingly rapid. By the 
time that some fifty reports (selected to give an approximate cross-section 
of the company’s interest) had been processed, more than 80 per cent of all 
the descriptors in the final schedule had been found. 

In this stage of developing their system, the panel had many lively dis¬ 
cussions about the theory and practice of using descriptors in information 
retrieval. These discussions were encouraged by the Zator representative, 
and the various points raised provided an excellent opportunity to bring 
up the experiences of other Zator clients who had similar problems. 

At the second meeting of the panel, about a week after the first meeting, 
the descriptors obtained so far were written down on a large sheet. This was 
the first draft of the descriptor schedule (see Table 15-2). Related descrip¬ 
tors were grouped together in the draft; duplicate descriptors were elimi¬ 
nated. 

Additional reports were then analyzed in the same way. Now the panel 
began to use the draft schedule as a guide. A few more descriptors were 
added, and rough spots in the draft schedule were ironed out. During this 
stage, scope notes were being written on index cards (for later typing in 
list form). Decisions made by members of the panel about the usage of the 
descriptors were thus written down while the problem was fresh ffi their 
minds. At various times the Zator representative would ask questions or 
offer criticisms to make sure that the panel was aware of the consequences 
of their decisions. Except for the teaching and guidance of the Zator repre¬ 
sentative, the panel did all the work in deriving their schedule of descrip¬ 
tors. 

The panel at Allied Research put in a total of less than 150 man-hours 
from the beginning of the process until the schedule was ready to hand over 
to their clerical staff for typing and code assignment. This time included the 
“homework” that was assigned to the various panel numbers between 
visits of the Zator representative. 

During the entire operation of deriving the descriptors, it was stressed 
that the primary orientation of a retrieval system mpslf be toward the 
requirements of the user. One of the most important consequences of user 
orientation is that the descriptors must be broad in meaning. When the 
descriptors are broad, the user’s intellectual universe can be covered by a 
relatively small list of descriptors. At Allied Research, 250 descriptors are 
used. Because there are so few descriptors in the system, they are relatively 
easy to remember, which is a definite advantage. The very breadth of 
meaning of each descriptor makes it easy to decide its applicability to a 
given document. Descriptors with finely drawn distinctions between them 
are avoided. Precision is not lost by using broad descriptors because ideas 
can always be synthesized by means of several descriptors. With so few 



350 


PUNCHED CARDS 


descriptors, it is easy to set them down on one big sheet, called the descrip¬ 
tor schedule. In this way, the analyst is able to see all the descriptors at 
once. 

At Allied Research, the booklet containing the scope notes (alphabetically 
arranged and printed), and the large sheet which is the descriptor schedule, 
are distributed to the engineers who are most active in using the system 
or who are on the team analyzing the incoming reports. To aid further in 
finding the correct descriptors, the scope notes have interpolated words 
and expressions in ordinary technical usage with cross-references to the 
proper descriptor. 

Analysis of the Incoming Documents—The Filtering Technique 

When the incoming reports arrive at Allied Research’s document center, 
they are given a preliminary screening to determine which engineer analyst 
is to handle each report. About sixteen engineers and scientists are on the 
analytical team. Each person gets the reports most closely related to his 
specialty. This procedure has the added advantage that it also keeps the 
specialists cognizant of the latest work in their fields. 

The procedures adopted for document analysis are also user oriented. 
No attempt is made in analysis to code the message of the document by 
writing a little abstract using descriptor words. The descriptors and their 
codes are used for retrieval only, and the message itself will always be 
available in the document. Neither is there an attempt to secure pin point 
precision with the descriptors. Excessively narrow descriptors will only 
frustrate the user when he attempts retrieval. 

The user of a retrieval system has a difficult problem. He is confronted 
by nothing but a schedule of descriptors supplemented by the scope notes. 
He is not sure what the file contains. He frequently knows nothing about 
the finer details in the reports. Thus, with only the schedule and scope 
notes, he must be able to formulate a prescription that will retrieve informa¬ 
tion, the nature of which may in large part be unknown to him. His suc¬ 
cess will depend largely upon how well the analysts originally did their job. 

In conformity to the philosophy of user orientation, the document analyst 
is asked to place himself in the user’s position. He does so in this way. First 
he reads or skims over the document. Then he lays the document aside and 
concentrates upon the descriptor schedule. He works down the schedule, 
descriptor by descriptor exactly as if it were a check list. For each descrip¬ 
tor he asks, “Would anyone at Allied Research who is interested in the 
content of this document use this descriptor as a part of his retrieval 
prescription?” or, “Does the meaning of this descriptor touch in any way 
upon the message of the document?” If the answer is “yes” to any of these, 
the descriptor is chosen as one of those to characterize the document. 



ZATOCODING INFORMATION RETRIEVAL SYSTEM 


351 


This is known as a “filtering” technique and according to this technique, 
the schedule of descriptors is filtered through the message of the document. 
Those that remain in the filter are the chosen descriptors. If there is any 
doubt about the applicability of any descriptor it is resolved by choosing 
the descriptor. A doubtful descriptor may be just the one that will be tried 
in a retrieval prescription by some eventual user. 

This technique has proved to be invaluable in giving the retrieval system 
a consistent intellectual structure. Consistency is a real problem. There 
are as many as sixteen or more contributing analysts at Allied Research 
and this group continues to change over the years. Yet their efforts ac¬ 
cumulate in the form of the Zatocard collection. These cards must be con¬ 
sistent to be usable, and rigorous application of the filtering technique has 
forced internal consistency of the system. 

The filtering technique also has another advantage. It does not require 
the analyst to have a highly technical background. If there were no filter¬ 
ing method, heavy demands would be placed on his ability and imagination. 
He would have to foresee all the possible uses of the document in order to 
decide which descriptors would apply. This is very difficult. However, the 
schedule of descriptors almost eliminates this problem because it serves as 
a check list of present and future contingencies as worked out by the top 
people in the laboratory. When the analyst uses the schedule as a check 
list, he only has to make very simple decisions. 

The burden of using a schedule of 250 descriptors is eased by a simple 
process. About one-quarter of Allied Research’s descriptor schedule is 
shown in Table 15-2. The descriptors are grouped, with each group being 
composed of similar descriptors. At the top of each of the groups there is a 
question, such as, “Is there a type of fluid flow?” In using this kind of a 
schedule, the analyst first looks at the questions. If the answer to any 
of them is “yes,” then he picks out the one or more appropriate descriptors 
below the question. If the answer is “no,” he continues to the next question. 
Use of the filtering technique in the Zatocoding dictionary system involves 
going through a list of about 20 questions rather than through 250 in¬ 
dividual descriptors. Carefully chosen “leading” questions, as in this exam¬ 
ple, make the incoming document analysis particularly easy. 

This grouping of descriptors is not a scheme of hierarchal classification. 
There are no generic or specific terms. Any descriptor can be used with any 
other, and more than one descriptor from a single group can be used to 
characterize a document. A typical document in Allied Research’s collec¬ 
tion has from six to fifteen descriptors in its characterization. It is some¬ 
times convenient to place the same descriptor in two different groups. This 
is useful for a few of the descriptors that may appear in widely differing 
contexts. An alphabetically arranged list of descriptors is specifically not 



352 


PUNCHED CARDS 


used for the analysis of documents because of its inferiority to the grouped 
arrangement in providing accurate analysis. 

An actual analysis proceeds as follows. The first decision of the analyst 
is whether or not to include the document. Obviously worthless material 
must not be allowed to increase costs or to dilute the system. By the time 
the analyst sees the document, it has passed this threshold of utility. The 
analyst then skims or reads the document. Depending upon the obscurity 
of the writing, or the richness of the content (there is often an inverse corre¬ 
lation), this usually takes from 5 to 25 minutes. Fifteen minutes is not a 
pessimistic average for technical reports. The analyst then takes the de¬ 
scriptor schedule and reviews the check list of questions, writing down the 
chosen descriptor words on a Zatocard. This step of filtering and writing 
down the descriptors takes about two minutes. The card then goes to the 
clerical staff who types or writes the title, authors’ names, report file num¬ 
ber, and any similar information on it. To save time and expense, the ab¬ 
stracts are not typed in. The clerical staff then marks the cards and notches 
them with the descriptor codes. The finished cards are not kept in any 
particular order. The documents are filed by number, and the process is 
complete. 

Most of the analyst’s time is taken in becoming familiar with the docu¬ 
ment. The Zatocoding System is not unusual in this respect. Regardless of 
the system used, comparable time will be required if the documents are to 
be analyzed to a like degree. The assimilation step accounts for 50 to 75 
per cent of the total cost of operating a retrieval system, with clerical costs 
and overhead accounting for the rest. 

Since the card services of ASTIA, AEC, and NACA are available to 
Allied Research for a large fraction of their reports, they are able to use a 
“clip and paste technique” to save typing. The stock used for Zatocards is 
heavy enough to support the weight of pasted-on material and it does not 
interfere with the sorting. Everything except the citation and the abstract 
is trimmed off the catalog cards before pasting. When such a Zatocard is 
sorted out, it carries the full printed abstract of the report. 

Coding the Cards Accurately 

Since Zatocoding uses random-like code patterns for the descriptors, and 
since random patterns are difficult to remember and to transfer accurately, 
there is a serious problem in coding. Its complete solution lies in the elimina¬ 
tion of the mental transfer step. The technique is illustrated in Figure 15-5 
which shows one page of a code pattern dictionary with a card in position 
for coding. 

To use this code pattern dictionary, the clerk reads a descriptor from the 
card, finds the page and line of the descriptor, and lays the card down on the 



ZATOCODING INFORMATION RETRIEVAL SYSTEM 


353 



Figure 15-5. The code pattern dictionary in use for transferring the code pattern 
for the descriptor “downwash” to the Zatocard. With the card in this one position, 
the location of the code notches for both the top and bottom edges are marked with 
a pencil. The cards are punched later. 

page under the descriptor entry. The first notching position of the card 
(site number one) is aligned with the vertical index line on the page. With 
the card in this position, the V-shaped marks on the page indicate the exact 
locations at the top and bottom of the card that are to be notched. The 
clerk transfers the code positions to the card with pencil marks. There is no 
error-prone mental step, so the accuracy is high. After the codes for all the 



















354 


PUNCHED CARDS 


descriptors have been made, the marked sites are notched with a hand 
“ticket punch.” 

Scatter Coding for Author and Company Names 

The last component of the dictionary system deals with authors’ names, 
company names, and the like. Instead of setting these up as individual 
descriptors and explicitly assigning them code patterns, a technique of 
scatter code ciphering is used to produce random-like patterns. By using 
such a ciphering process, a vast number of assigned patterns for little- 
used names is avoided. Scatter coding is primarily used for author names. 
When there is more than one author, the additional authors are ciphered 
into the card together with the first. 

As shown in Figure 15-6, a card to be coded with an author’s name is 
laid down on the scatter code sheet with the left-hand edge of the card op¬ 
posite the index “N” (for name). The first four letters of the surname are 
then spelled out. Two letters from the name are ciphered at the top of the 
card and two at the bottom. Notice that the letters of the two alphabets at 
each edge are displaced so that letters of high frequency like “e” or “t” 
do not coincide at the same site on the card. This kind of displacement of 
the alphabet insures that the scatter-coded entries will have notches with 
an approximate uniformity of incidence of notches across the edge of the 
card as required by the Zatocoding method. The scatter codes are suffi- 

SCATTER. CODING SHEET 



Figure 15-6. Diagram illustrating the ciphering of an author’s name by scatter 
coding. 








ZATOCODING INFORMATION RETRIEVAL SYSTEM 


355 


ciently random-like so that the usual number of names on any card will 
not upset the sorting statistics. Standard rules are employed with company 
names to eliminate parts of the name useless for sorting, like “corporation.” 
Such rules enable the elimination to be done in a consistent and repeatable 
fashion. This particular version of scatter coding was chosen for use at 
Allied Research after several other schemes had been considered. 

The Need for Rapid Cyclic Search—Machine Feedback 
of Information 

The Zatocoding system at Allied Research is actively used by the en¬ 
gineers for a variety of reference problems. Despite their excellent familiar¬ 
ity with the system, and their background in the various technical fields 
covered, it frequently happens that an engineer making a search is unable 
at first to prescribe exactly what he wants from the document file. He has 
to do some browsing before he knows how to sharpen his question so as to /' 
obtain the best answer. The way he does this is to formulate the best trial 
prescription that he can. He then sorts the cards and looks over the selec¬ 
tions. He goes through the titles and abstracts to see how close he came to 
what he thought he wanted. On the basis of this preliminary work, several 
things may happen. He may find exactly what he wanted in the way of 
information, or he may actually change his mind as to what he does want. 

A third possibility is that he may decide how better to prescribe his search. 
Thus he may omit some of the descriptors from his prescription and add 
a few others. With such a new search prescription, he is ready to make 
another search. He may make a second or even a third search. From each 
search he learns more about the content of the file and how to ask his ques¬ 
tions to match his technical problem. 

A cyclic search process is unavoidable in creative science and engineering / 
because the questions that arise are often diffuse and the details of the 
looked-for facts and theories are at first unknown to the searcher. This is 
why a recourse to more elaborate coding or information analysis cannot 
eliminate this problem. The shortcoming lies in the user and not the re¬ 
trieval system, and it is one of the functions of the system to educate the 
user at the question-asking stage. To do so, there must be a rapid feed back 
of corrective information. At Allied Research the cards can be completely 
searched in six or seven minutes, and they will provide an immediate 
answer in the way of titles or abstract. 

In many cases particularly with diffuse questions, the occurrence of 
extra cards is helpful to the user. For instance, in making a selection from 
descriptors A, B, and C a few extra cards’will often come from the selector 
with only descriptors AB, or AC, or BC of the original prescription. One 
descriptor is missing. These few extra cards have a valuable property in 



356 


PUNCHED CARDS 


that they give a random cross-section of information outside the original 
search prescription, but a selection that is still heavily biased in the direc¬ 
tion of the search prescription. These cards have been called “subject in¬ 
duced extra cards.” 

As the cards drop from the selector, they have the title and often the 
abstract of the report in plain language. The cards also have all their de¬ 
scriptors written in. Thus when an interesting technical lead appears, it 
is immediately evident which descriptors should be used in following it. 

An Evaluation 

The Zatocoding System at Allied Research Associates, Inc. has been in 
operation for nearly four years (August 1958), and contains about 8,500 
documents. The whole file can be searched in twenty minutes. New docu¬ 
ments are added at a rate of about 150 per month. Searches are made on the 
average of once a day. Any engineer, after a short training period, can 
analyze documents and make searches. Two young women are working full 
time running the document center and coding and notching the cards. The 
system at Allied Research is generally accepted as a straight-forward work¬ 
ing tool, quite in the same way that a desk computing machine is accepted 
and used. In seeking contracts, the company now stresses its outstanding 
ability to retrieve information. 

The system has worked smoothly and achieved the expected performance. 
During the period of installation, the advice of the Zator representative 
was helpful in foreseeing pitfalls in advance. There was no substantial 
backtracking nor need to correct mistakes. 

As a result of experience, about a dozen descriptors were added to the 
dictionary during the first half-year, and about the same number were 
dropped. Since then the coding dictionary has had considerable stability. 

The costs for this commercial system are: $45.00 monthly rental and 
license for the Zator card sorter and for the Zatocoding technique, $15.00 
per thousand for cards, a professional fee for assistance in installation, and 
traveling expenses incurred by the Zator representative. 

In just a year, the technical files were smoothly transformed into an 
efficient operation which is an engineering asset to the company. 



CHAPTER 16 

THE USE OF PUNCHED CARDS IN 
LINGUISTIC ANALYSIS 


Rev. Roberto Busa, S.J. 

Centro per l’Automazione dell’Analisi Letleraria dell’Aloistanum 

Gallarate, Italy 

A chapter describing the application of punched cards to studies in the 
humanities has a place in a book such as this, because such application in 
part parallels, and in part coincides fully with its corresponding use in sci¬ 
entific documentation and in libraries. I maintain that it is mutually ad¬ 
vantageous to consider how the same tool for investigation responds to the 
demands of many problems differing in their nature. 

I am concerned here with “linguistic analysis” in a broad sense, rather 
than in any of the specific meanings that different schools have sought to 
impose upon the phrase. I refer to any type of investigation of language, 
whatever significance the word “language” can assume. For example, I in¬ 
clude the study of phonetics, of glottology, grammar, or style. In a word, I 
speak of philology in its broadest sense, and of psychology. I speak only 
of the investigation of written material, or more strictly, printed words. 
Even studies of phonetics can be based on printed texts. Hence, I am not 
concerned with those other analyses dealing directly with human sounds, 
such as those conducted at the Haskins Laboratories 1 in New York, nor 
those conducted with devices on which data are not recorded in letters or 
symbols (e.g., studies of comparative phonetics). 

The analysis of language is as old as the knowledge of human knowledge. 

1 These studies are aimed primarily at isolating the significant signals embedded 
in the speech stream and in analyzing their perception by the human listener. An 
analysis-synthesis technique is used wherein the speech is converted into visible 
patterns, the patterns are re-drawn in simplified form and, finally, the modified pat¬ 
terns are re-converted into synthetic speech to provide the acoustic stimuli for 
perceptual studies. 

“Some Results of Research on Speech Perception,” A. M. Liberman, The J. 
Acoust. Soc. Amer., 29, No. 1,117-123 (1957). 

“The Interconversion of Audible And Visible Patterns As A Basis For Research 
In The Perception Of Speech,” F. S. Cooper, A. M. Liberman, and J. M. Borst, 
Proc. Natl. Acad. Set., 37, No. 5, 318-325 (1951). 

“Some Experiments on the Perception of Synthetic Speech Sounds,” F. S. Cooper, 
P. C. Delattre, A. M. Liberman, J. M. Borst, and L. J. Gerstman, The J. Acoust. 
Soc. Amer., 24, No. 6, 597-606 (1952). 


357 



358 


PUNCHED CARDS 


Even without disturbing Plato in his Dialogues, it would be necessary only 
to recall to mind the rhyming dictionaries and the hundreds of concord¬ 
ances that have been published since the invention of printing. In more 
recent times, there has been increased interest in literary statistics. Refer¬ 
ence was made to quantitative statistical analyses in formulating psycho¬ 
logical and stylistic laws, for example, on the length of phrases, on the 
distribution of phonetic accents, on absolute and comparative frequency of 
words, of parts of speech, or of phonemes (in the meaning of letters of the 
alphabet). There are scholars in the United States who have made important 
progress in this field. 2 

Therefore the subject matter to be analyzed is made up entirely of what 
can be transcribed from human speech into characteristic signs or symbols. 
It can be considered as having three levels. First of all, the word is the 
fundamental unit, and it is at the same time the graphic and semantic unit. 
Then there are sentences and phrases composed of more than one word. On 
the other hand, there are elements of each word, such as roots, prefixes 
and suffixes. In the same way we speak of atoms, molecules, and electrons. 

Thus it would be interesting to know which are the words used by a person 
or an epoch or a language. How many are there? To what radicals can they 
be reduced? What is their frequency? their length? What are the rhythms 
of their accents? How are words distributed in phrases? What are the 
fundamental structures common to the phrases? There are many such 
questions. 

The problem requires the searching, separating, arranging, correlating 
and study of a large number of small elements, tens of thousands of words, 
hundreds of thousands of letters. Such investigations must be repeated 
many times on the same material with different emphases and for diverse 
purposes. 

For these studies we must record every unit of information on a free and 
manageable medium, such as a card. Punched cards permit multiple coding 
of the same information, and they can be sorted and re-sorted rapidly. In 
addition, the great—even enormous—quantity of cards to be handled, and 
the possibility of making automatic printouts directly from the cards, dic¬ 
tated the choice of machine-sorted punched cards. Among these I have 
finally chosen the IBM system, not only because Providence obtained for 
me the full collaboration of this company, but also because of the great 
flexibility of the system, because of the developments the company foresees 

* For a list of U. S. scientists in this field, see for example Guiraud, Pierre. Bib¬ 
liographic critique de la statistique linguistique. Revis4e et compl4t6e par Thomas 
D. Houchin, Jaan Puhvel, et Calvert W. Watkins, sous la direction de Joshua What- 
mough. Utrecht, Editions Spectrum, 1954. 

xix, 121 p. (Comit4 international permanent de linguistes. Publications du Comit6 
de la statistique linguistique, 2). 



USE OF PUNCHED CARDS IN LINGUISTIC ANALYSIS 


359 


Original text 
(Marked by scholar) 

> 


Sentence cards punched 
(Clerical key-punching) 

(See Figure 16-2) 


• 

Word card prepared 
(Automatic processing) 

(See Figure 16-3) 



Form cards 
and 

Main cards 

1- 

-1 


Concordance and other 
Listings for 
linguistic analysis 
(See Figure 16-4) 


Figure 16-1. Summary of operations. 

in the near future, and because IBM has been developing machine methods 
for scientific documentation. 

I will now recount all that I have done, and all that there is still to do. 
I will use a flow chart (Figure 16-1) as the basis for my discussion. This was 
first prepared at IBM in Milan by Mr. C. Folpini and then completed in 
the offices of IBM in New York with the assistance of Mr. P. Tasman.* It 

* Literary Data Processing,” P. Tasman, IBM J. Research & Development, 1, No. 
3 , 249-256 (1957); 

“Literature and Document Research Automation,” P. Tasman, Automation Sys¬ 
tems, Engineering Publishers Division of The AC Book Company, Inc., 1958; 61-72. 








360 


PUNCHED CARDS 


will be evident that the process so described could be shortened considerably, 
if it is sufficient to obtain simpler results. However, it is useful to give the 
whole picture to show how much can be obtained if desirable. 


Analysis of Words in a Text 


I will concern myself here with the principal task, in terms of size and 
value, of linguistic investigation, which is making a concordance of a con¬ 
tinuous text. It will not be difficult to apply these techniques to other prob¬ 
lems, such as analyzing the answers of questionnaires, or the words found 
as items in a glossary. 

The scholar marks the text to indicate how it should be recorded on the 
cards, noting the beginning and end of paragraphs and of sentences with 
their appropriate references. Also he distinguishes words quoted by the 
author from other writers, from the author’s own words, etc. 

Where it is important not to mark directly on the text so as to deface it, 
a sheet of cellophane may be placed over the page and appropriate mark¬ 
ings made on it with washable inks. 

Each line from the text is punched into a card, one line after the other, 
each with identifying reference to its place in the text. The maximum num¬ 
ber of columns available for this punching is determined in advance, de¬ 
pending on the format wanted for the concordance. Thus a maximum 
sentence length is established. Words are never split between cards; rather, 
a word is started on a new card if it will not fit on the preceding one. See 
Figure 16 - 2 . 

The problem of verifying the punching is an important one, because an 
undetected error will always be repeated. Errors can be detected in the 
usual way by checking the card on the verifier, or by proofreading cards 
on which the punching has been interpreted. 


PW7U3 Esse* m rwincirta |ti» ni^i* 

H >• T I 111* I II I M I « 


|t I'M! I' I e 1 9 c |||ritiYiiTiTmirri«miit11| ||illij1 1 |:; t ? t •:»i * • •:»t o o o o 

i|i fill 11 |i' |i i m 11 n m hi 1111111 1 ; i |i 1 1 1111 111 n 11 111 1 ii 111 m 1111 * < 

tiiiiimtiimi|iiiiimii!itmm : miiii|j|.'i!itmiiiiiiiimiui 2 2 2 2 

j 111 ttin 111 i!*» 111111 J 1 } 11111 ii|i 1 1 *1 1 J 15 ii''■ 11 1)>‘jj 11 »iJ|i 1111 > 111»»11 5 "Tf 

<444|44|««444 4 |4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 |M 4 | 4 4 4 4 4 4 I 4 4 4 4 4 4 M 4 4 4 4 4 4 I 4 4 * 4444 

]ih|nm»|iii I I I I 

| rnmiTTOfim nmi:?) ri|imi|: fiiiiriniijnnKinninmnn^ ^ * ? 

* Ull|l|4llt»ll«l MimtMlf III III lli;l««l« IfMtlflll I»II|»«SI|I11II91 f 0 ° * 8 


Vf tSM49ft*H«| 


H 1ft HI MJ Vt II ff9* iff ’ lift f ft** 

• ■ " w , T n T. ^ *1 !f w ft- 1 .T 


Figure 16-2. Sentence card. 




USE OF PUNCHED CARDS IN LINGUISTIC ANALYSIS 


361 


These operations produce the first or fundamental group of cards called 
the text cards, or on the flow chart, the sentence cards. With this first and 
only data transcription, it is possible to accomplish mechanically, speedily, 
and accurately all of the most diverse and complex analyses. 

To divide the sentences into single words, each on a separate card, the 
following methods could be used: 

(a) punching each word from the text onto a separate card. 

(b) simultaneous use of the sorter and reproducer equipment as de¬ 
scribed in the small volume, Varia Specimina Concordantiarum * 

(c) using the Cardatype, recently developed by IBM. 

The use of the Cardatype offers the advantage of preparing typed copy 
of the text while punching the individual word cards. Thus the context of 
the word is printed on the reverse side of each card. 

This operation results in a second set of cards, the word cards. Each word 
is accompanied by reference to its place in the text. This file contains as 
many cards as there are words in the text. See Figure 16-3. 

The word cards are alphabetized, using the sorter. Mechanical alpha¬ 
betizing requires two passes of the cards through the machine for each col¬ 
umn sorted. Thus sorting 100,000 cards containing words of 10 letters 
means, in effect, passing 2,000,000 cards through the machine. Depending 
on the model sorter used, from 30,000 to 60,000 cards per hour can be 
sorted. Therefore it could take from 35 to 65 hours, approximately, to ac¬ 
complish the alphabetization. In other words, the machine would alpha¬ 
betize from 1,500 to 3,000 words of 10 letters in an hour. 

The operation can be shortened by several means, e.g., by separating 
first the shortest (one- and two-letter) words, then the next shortest words, 
and so forth. The shorter sets so divided are then inserted into the alpha¬ 
betically sorted sets of longer words. The final result will be that all the 
words of the text are alphabetized and all identical words are grouped to¬ 
gether. 

Each group of identical words is given the same sequence number. 

The accounting machine is used to print a list of all the words from the 
word cards. It is possible to prepare an abridged list on which only the dif¬ 
ferent words appear. The machine will also print the total number of cards 
on which each different word appears. This gives the frequencies with which 
each different word appears in the text. The accounting machine may also 
be set to print the code number identifying the different forms of individual 
words. 

When the summary punch is connected to the accounting machine, a 
third series of cards can be obtained while the list described above is being 
printed. These cards, called form cards or different word cards, contain each 

4 Roberto Busa, Varia Specimina Concordantiarum, Fratelli Bocca, Milano, 
1951, 180 pp. 




362 


Figure 16-3. Example set of word car 






















USE OF PUNCHED CARDS IN LINGUISTIC ANALYSIS 


363 


different word with a number indicating its position in the alphabetic se¬ 
quence and the total frequency of its appearance. Such a group of cards is 
not necessary for ordinary concordance work but it may open up the pos¬ 
sibility of different and new investigations. Such cards, in fact, contain the 
summary of the author’s vocabulary and can be analyzed indefinitely. Mark 
sensing techniques are particularly useful for this (as will be discussed later). 

In the list of words described above, the words are considered according 
to their graphic structure. Therefore the scholars must separate cases of 
homographs, which turn out to be quite frequent; dismember words that 
comprise prefixes and suffixes each having a proper function also when 
isolated (such words may be considered as two words rather than one); as¬ 
semble the separate words that are in reality just one verb form; and 
finally, regroup under the functional semantic unit all the diverse forms a 
word assumes according to case, tense, mode, etc. 

Such work requires the competent responsibility of the scholar and it 
cannot be accomplished by machine. However, once such classification has 
been made, mechanical recognition of different forms of the same word 
could follow. 

The main words must be punched one per card with a special layout de¬ 
signed to accomplish the functions of these cards. They must also be ar¬ 
ranged in alphabetical order and numbered progressively. Such a number- 
code may be added to the cards of the other two sets, word cards and form 
cards. 

In the form that I have summarized there is obtained from one initial 
punching of the text, four groups of cards. They are: the text cards and the 
word cards, that contain all the words of the text and represent two new 
editions of the entire text; and the form cards and the main cards, which 
constitute two summary indexes of the vocabulary used in the text. The 
first lists the words grouped according to graphic form, the other lists the 
same words arranged according to graphic-semantic units. 

Note that the word cards are accompanied by the elements necessary to 
characterize their individuality. The dissociation of the text into its first 
elements is, therefore, entirely reversible: it is always possible to reconsti¬ 
tute the text from these elements. Such proper determinations, reserved 
and exclusive of each single word, are its various codes. In fact, every word 
is coded as to its location with the reference and with the number of its 
position in the text; it is coded as a morphologic unit with the progressive 
number that it acquires in the first alphabetic sequence; it is coded as a 
semantic unit, with the progressive number that it has in the last alpha¬ 
betical order. 

Besides, it is accompanied by its context. Such context may be punched 
or printed. It may be punched on the same card on which the word is 



364 


PUNCHED CARDS 


punched; a condition, however, that is restricted to about 50 to 60 letters 
and spaces, i.e., columns on the cards. It may be punched on another card, 
and then occupy even 70 or more columns, according to the length of the 
reference; such a card would then be the same text card. The context may 
be also printed in the spaces between the punched holes, and then it can 
be extended to twelve lines, and contain from 80 to 120 words; ample con¬ 
text that would almost always be sufficient to individualize the significance 
of the word without requiring the scholar to make frequent recourses to 
the printed text. 

Finally, every single word can be accompanied by the first or last letter 
of the preceding word, by the first or last letter of the following word, be¬ 
sides the preceding quotation and the following quotation. 

There also remains in the word card or in the form card or in the main 
card sufficient space for additional classifications to be applied, for example, 
manually by means of mark sensing. The researcher can make a symbol 
that tells what part of speech the word is, on what syllable the tonic ac¬ 
cent goes, what is its length in letters or in syllables, and other things, too. 

So the resolution of the text in its first elements is completed. The four 
groups of cards represent the material suitable for whatever investigation 
in whatever direction: investigation that, in its quantitative aspect in¬ 
volving large numbers of small elements, is accelerated enormously by 
mechanization and rendered more accurate, more certain, and absolutely 
complete. The same cards will serve the most diverse analyses, because 
once the cards are selected according to a determined order, they may be 
brought back rapidly to their first order and subjected to new research. 

One can obtain, from the one and initial punching of the text, various 
listings as summarized below and exemplified in Figure 16-4. 

(1) The general catalog of vocabulary of the author, richer in preroga¬ 
tives and more abundant of context than the same monumental concordance 
of TLL 6 prepared in Munich, Bavaria. 

(2) The listing of the cards in various forms, using the accounting ma¬ 
chine at a rate of 4500 to 9000 lines/hour. For example: (a) The text cards 
may give a reprinting of the entire text, (b) The word cards may give the 

6 TLL— Thesaurus linguae latinae, editus auctoritate et consilio Academiarum 
quinque Germanicarum Berolinensis, Gottingensis, Lipsiensis, Monacensis, Vindo- 
bonensis. Lipsiae, Teubner, 1900. 

Current volumes read: Thesaurus linguae latinae, editus iussu et auctoritate 
consilii ab academiis societatibusque diversarum nationum electi. 

“The great dictionary of the language, in Latin, indispensable in the university 
or large reference library. Plans to record, with representative quotations from 
each author, every word in the text of each Latin author down to the Antonines, 
with a selection of important passages from the works of all writers to the seventh 
century.” Winchell, Constance M. Guide to reference books. 7th ed. 



LATERCULUM VERBORUM 

Numerorum qui singula subsequentur verga primus fro- 
quentiam , alter cui lemmati in Rationario adunetur 


indieabit. 

1 A 2 1 

2 AB 11 

3 ACCIPITE 1 2 

4 AD 4 3 

5 AEMULIS 1 4 

6 AGITUR 1 6 

7 AGNUM 2 5 

8 AGNUS 1 5 

9 AMBIGITUR 1 7 

10 ANGELICU8 1 8 

11 ANGELORUM 1 9 

12 AN1MOSA 1 10 

13 ANTIQUUM 1 11 

14 A88UMITUR 1 12 

15 AUDE 1 13 

16 AUXILIUM 1 14 


A. Alphabetical listing of words as they appeared in 
the text. (Note that a serial number precedes the 
word as listed. First number after each word indi- 


CONSPECTUS LEMMATUM RATIONARII 

1 A AB 

2 ACCIPIO ACCIPI8 ACCIPERE 

3 AD 

4 AEMULUS AEMULA AEUULUM 

5 AGNUS AGNI 

9 AGO AGIS AGERE 

7 AUBIGO AMBIGI8 AMBIGERE 

8 ANGELICUS ANGELICA ANGELICUM 

9 ANGELUS ANGELI 

10 ANIMOSU8 ANIMOSA ANIMOSUM 

11 ANTIQUUS ANTIQUA ANTIQUUM 

12 A8SUMO ASSUMIS A88UMERE 

13 AUDEO AUDES AUDERE 

14 AUXILIUM AUXILII 

15 AZYMUS AZYMA AZYMUM 

16 BELLUM BELLI 

17 BENEDICTIO BENEDICTIONIS 

18 BIBO BIBI8 BIBERE 

B. Main word listing. (Note seQuential number of 
these main words and citation of various 
forms.) 


cates frequency of appearance in text and the 
second number refers to the “Main word listing.”) 

RATIONARJUM VERBORUM 

Poet singula lemmata proprio numerata numero, vo- 
cabula singula numerus praecedei quo in Laterculo 
prime continebantur, numerus vero subsequetur fre- 
quenliae singularis ac tandem collectivae. 

1 A AB 

1 A 2 

2 AB 1 

3 

2 ACCIPIO ACCIPIS ACCIPERE 

3 ACCIPITE 1 

1 

3 AD 

4 AD 4 

4 

4 AEMULUS AEMULA AEMULUM 

5 AEMULI8 1 

1 

5 AGNUS AGNI 

7 AGNUM 2 

C. Word index combining all A-words with oorres- 


INDEX VERBORUM 

1 A AB 

A L 

A V 

AB p 

2 ACCIPIO ACCIPIS ACCIPERE 

ACCIPITE 8 

3 AD 

AD P 

AD S 

AD V 

AD y 

4 AEMULU8 AEMULA AEMU¬ 
LUM AEMULIS V 

5 AGNU8 AGNI 

AGNUM 8 

AGNUM 8 

AGNU8 8 

6 AGO AGIS AGERE 

AGITUR L 


15 43 

2 5 

6 35 

4 15 

4 23 

7 28 

1 3 

1 4 

2 6 

2 6 

3 9 

21 66 

6 16 


D. Word index combining word entries from A and 
B with citation of all occurrences in the text. 
(Numbers to left refer to listing B.) 


ponding B-words indicating—to right—frequency 
of occurrence. (Numbers to left refer to listing B.) 


CONCORDANCE 


1 A AB 

A 8UMENTE NON CONCISU8 
L 15 43 

A 

PROCEDENTI AB UTROQUE 

P 6 35 

AB 

IN MORTEM A DISCIPULO 

V 2 5 

A 

2 ACCIPIO ACCIPI8 ACCIPERE 

DICENS ACCIPITE QUOD TRADO 

VASCULUM 8 4 15 

ACCIPITE 

3 AD 



AD FIRMANDUM COR SINCERUM 
P 4 23 AD 

AD LUCEM QUAM INHABITA8 

8 7 28 AD 

AD OPUS 8UUM EXIENS 

V 1 3 AD 

VENIT AD VITAE VESPERAM 

V 1 4 AD 


E. Concordance. (Listing of main words in textual 
context with identification of location in text. 
Numbers to left refer to listing B.) 


Figure 16-4. Concordance and other listings. 


365 



366 


PUNCHED CARDS 


alphabetic list of all of the different forms under which the words used are 
presented in this text, indicating their frequency. This laterculum formarum 
may be obtained immediately after the words have been alphabetized and 
arranged. But if the code number of the main word is required, it would 
be necessary to wait until after the rationarium verborum has been prepared, 
(c) The rationarium verborum or formarum would be the diagram, systema¬ 
tized and with frequencies, of all the same words regrouped according to 
their meaning, or more exactly, according to the identity of their functional 
elements. Such a list is the basis of the author’s vocabulary, (d) It would 
be very simple to list an abridged conspectus lemmatum. (e) The index 
verborum will be the index of all the words, or rather of all the word cards, 
with the reference and arranged according to the rationarium. (f) The Con¬ 
cordance will be the same list with the single words followed by the nota¬ 
tion as well as the reference. The context may be of one line only, what¬ 
ever is deemed to be sufficient; but it may also consist of three or more 
lines; in this case the word in question will always be found in the middle 
line. 

For a simple Concordance, preceded by the laterculum and rationarium 
formarum, the required time will, for the first phase, equal the time of one 
or two typings of the entire text; for the following phases, it will be possible 
to fulfill in one year that which would take 30 to 40 years of work with the 
old method. This is the case for the printed Concordance. When, however, 
one needs to compose a catalog in which the words follow a context of 12 
lines on single word cards, 20 or 30 years work can be completed in one year. 

In respect to the cost of the work, this much was made clear, on the basis 
of Italian prices rather than those in the United States: We compared on 
the one hand a form, composed correctly and paginated, ready to be put 
into the rotary press, and the on other hand a Concordance obtained from 
the IBM accounting machine on mats adapted for lithographing, and ready 
to be passed through the offset system or any other system of lithographic 
reproduction. We did not include the cost of materials, paper, or zinc. A line 
of the Concordance prepared and tabulated with the IBM system costs 
about half what it would cost to set up a line with a linotype or monotype. 
The computation was made on the supposition that all the work is done 
in the IBM offices at commercial prices. The difference in the cost will be 
more appreciable, if one realizes that the cost of conventional printing 
does not include the cost of preparing the Concordance; while the cost of 
the IBM listing also comprises all of the work and materials of preparation, 
such as punching, sorting, and reproducing the cards. In other words, the 
new method, at half the price required for the preparation of the printing 
of a Concordance, gives not only the matrices for printing, but also the 
entire catalog in a flexible form always ready for new studies. 



USE OF PUNCHED CARDS IN LINGUISTIC ANALYSIS 


367 


Research on the Structural Elements of Words 

A statistical list of prefixes is already contained in the first laterculum 
verborum. It is also very easy to obtain a list of only the first three or four 
or more letters of words, with totals of frequencies of the single different 
beginnings of words. It is evident, however, that in this case the machine 
will list also the short words composed of less than five letters, unless the 
scholar prevents this by appropriate instructions to the machine. 

It is possible to sort from the sets of word cards and form cards those 
words with particular combinations of initial four of five letters. One can 
also use the collator in which a pilot card punched with only those letters 
that constitute the prefixes desired, instructs the machine to extract those 
cards that contain that composition of initial letters. 

If cards are placed in the accounting machine to obtain the list of word 
cards or form cards, a summary punch can be coupled to obtain a series of 
cards that represent the various be ginning s of words, accompanied by a 
code number (a serial number representing the alphabetic order) and by 
the total of the frequencies. 

In like manner one can turn to the analysis of the endings. For this we 
need to punch the words so that the last letter of each word is in the same 
column. This can be done with the sorter, by separating the words by length, 
then reproducing all of the cards so that the last letter of each word will be 
found in the same column. This task is simplified by working from form 
cards. The words so punched are now alphabetized backwards. This is done 
by sorting first the initial letter of the longest word and then proceeding 
from left to right. (This in the reverse of the usual IBM alphabetizing 
procedure.) We now have the reverse index, in effect a rhyming dictionary. 
In this way we can list different endings of words indicating the number of 
frequencies. Also we can pair the accounting machine and the summary 
punch to obtain a series of cards, which contain the endings of the words, 
in order to work only on these. 

The calculation of the number of letters of a text or of a vocabulary be¬ 
comes a very simple operation when those words are already punched in 
cards. One can, for example, use the text cards and explore every single 
column to sort the letters present in that column. Also we can count each 
package with the card counter of the sorter, and write the sum totals. To 
the total of the first column we add the sum of the letters present in the 
second column and so on. This operation is facilitated by using a sorter 
provided with a counter, and even more by using the 101 statistical ma¬ 
chine. 

More work is necessary, but it is still extremely fast compared to manual 
labor, when one undertakes to analyze the distribution of letters, diph¬ 
thongs, for example, in words. Such an inquiry, in fact, coincides with the 



368 


PUNCHED CARDS 


search for roots of words. Such inquiries can be done by means of succes¬ 
sive sortings. This could be shorter if the collator were used, with a pilot 
card. It would be shorter yet if the 101 statistical machine were used: this 
machine searches at the same time four different groups of three letters in 
each word of twelve letters punched in a card. The cards pass through the 
machine at a rate of 46,000 per hour. The machine does not select the roots 
only, but rather any combinations of letters specified. 

The machine will also facilitate the preparation of materials for study 
of the distribution of tonic accents, of the proportion of use of the parts of 
speech and other things. For example: what per cent of nouns, of verbs, of 
adjectives ... or what is the predominant structure of the phrase: subject- 
verb-complement (or predicate), or instead, predicate (or complement)- 
verb-subject. 

Mechanical Search of Phrases 

The actual system as described permits searching for a particular phrase, 
if as emphasized above, the first letter of the preceding word and that of 
the following word were punched on the word cards. There is, for example, 
the saying, “sotto questo punto di vista.” Among all the cards that carry 
the word questo, the sorter separates those in which questo is preceded by s 
and followed by p. Among the words di, those are selected that are pre¬ 
ceded by p, preceded in turn by q and s, and followed by v. These cards 
give the references to all the passages in which are found said sentence, 
even if the sentence is distributed on two successive cards. 

It must be noted here that all that has been said is not necessarily limited 
to Latin characters. The machines can be provided with any alphabet, and 
also for Arabic and Hebrew which proceed from right to left. Any series of 
symbols, signs, or ciphers may be applied to the machines. 

The analysis requirements for most texts necessitates the use of punctua¬ 
tion and diacritical marks. The IBM accounting machines such as the 402, 
403 and 421 can be utilized for these marks by substituting for the Arabic 
numerals the desired symbols. This is also the case where card interpreta¬ 
tion is required on the IBM 552. Card punches may be modified by substi¬ 
tution of suitable key tops. 

For those symbols that should accompany the word, the 12 and 0 zone 
punch positions of the card should be reserved for use with all IBM ac¬ 
counting machines except the 407. For example, the apostrophe for the 
articles with elisions in Italian and for the genitive in English, and the point 
(or period) for the abbreviated words must accompany the word even when 
it is isolated in the word card. Thus, the German das ist becomes abbrevi¬ 
ated d. i., but is punched like d-i-; and is also listed as d-i-. 

Adoption of the diacritical symbols and of interpretations offers major 
possibilities through the use of the IBM 26 card punch and 407 accounting 



USE OF PUNCHED CARDS IN LINGUISTIC ANALYSIS 


369 


machine, where in addition to the numbers and the letters there are spaces 
for 11 special characters. Ordinarily such characters are accounting charac¬ 
ters, but they can with moderate cost be substituted by signs used in 
linguistic or philologic studies. 

Problems involving accents are more difficult to resolve by ordinary 
means. Some languages actually present a considerable problem: one thinks 
of modem French and classic Greek. When special characters for symbols 
of punctuation, diacritical marks and accents are required at the same 
time, the top space of a column is not sufficient, except for certain kinds of 
work and for some languages, for example English, Italian, or Latin. It is 
necessary then to consider IBM machines which use codes of punches con¬ 
sisting of combinations of two or more columns. 


The Near Future 

An application of punched cards that should be explored more fully is 
the automatic tracing of the variants of the same text. The first step in 
any critical analysis consists of comparing the results of the same analysis 
applied to a representative selection of various manuscripts or editions. 
The first line is written down, then the variants as encountered in the other 
copies. From this material the researcher determines whether the first word 
is an authentic word of the author. It would be possible to devise a process 
as follows: Punch line after line for all the editions judged to be representa¬ 
tive, one line per card with the reference and proper symbol for each 
edition. The sorter will assemble all the first lines according to the symbols 
for the editions, then all the second lines, then the third lines, and so on. 
The cards ordered in this way are fed to the accounting machine, set to 
print the group of first lines, then to leave a space before printing the group 
of second lines, etc. The machine can be so set that for every group, it 
prints the first line in its entirety, and only those parts of the following 
lines that are different from each preceding line. Probably it will also be 
possible for the machine to print, for all of the successive versions of the 

1 POCA FAVILLA GRAN FIAMMA SECONDA 

2 POCA FAVILLA GRAN FIAMMA SECONDA 

3 POCA FAVELLA GRAN FIAMMA SECONDA 

4 POCA FAVILLA GRAN FIAMMA SECONDA 

5 POCA FAVILLA GRAN FIAMMA ASSECONDA 

6 POCA FAVILLA GRAN FIAMMA SECONDA 

7 PRIMA FAVILLA GRAN FIAMMA SECONDA 

8 POCA FAVILLA GRAN FIAMMA SECONDA 

9 POCA FAVELLA GRAN FIAMMA SECONDA 

Figure 16-5. Tabulations of a Set of Hypothetical Variants of a Verse from Dante 

(Paradiso I, 34) 



370 


PUNCHED CARDS 


same line, only that which is different from the first line. Among the graphs 
is a model showing the tabulation for punching of a set of hypothetical 
variants of a verse from Dante. (Figure 16-5) 

Such a process presents problems of cost. It must be established whether 
the cost of punching and verifying the same text as many times as there are 
variations is compensated by the speed and certainty of the subsequent 
analysis and by the fact that the operation produces text cards for linguistic 
analysis and the preparation of a concordance of the text. 

There are technical problems yet unsolved. The principal problem has to 
do with prose writings. Due to the fact that the machine exercises control 
on each column of the card, should a single letter be left out of one fine as 
punched on one card, then the remainder of the text on other cards would 
be displaced by one column and would as a result be printed as a variant. 
Such difficulty does not exist for writings in verse, each line of which is 
started on a new card. For prose works the problem might be resolved by 
using punched tape or the electronic computer with sufficient memory ca¬ 
pacity. 

Another development is the printing of the differential context on the 
back of the word cards. If on all the cards for the words in a given section 
of text there is recorded the same context, then the first word does not have 
any preceding context and the last word does not have any following con¬ 
text. As the system is developed there is a need for the context to be printed 
in such a way that the word punched on the card will be found in the center 
line of those printed. Thus for all of the words of line 20, the text would 
begin on line 14 and end on line 25; for the words of line 21, on line 15 to 
line 26; for the words of line 22, from line 16 to line 27, and so on. This 
problem is not exactly one of machines for punching cards, but rather one 
of duplicating machines. However, it is desirable that the processes be linked 
with the punching on the same cards. 

In addition, two parts of the process need to be accelerated: reproduction 
from the text cards to the word cards, and alphabetic sorting. These two 
phases notably affect the time and cost of the analysis procedure. Difficult 
problems of these types obviously do not occur in the ordinary industrial 
and statistical applications of the IBM system. It is also obvious that, for 
this reason, an answer to such a demand will come in the future, when re¬ 
search work on linguistics has justified the cost of constructing new models 
of machines or at least adapting models already in use. 

The possibility of punching text cards automatically, starting with the 
examination of the text by means of a photocell or by other means exists 
but as yet there are no practical methods for carrying it out. Naturally it 
will be gratifying when such techniques are operational. 

For preparation of concordances, those most valuable kinds of philologic 
studies, perfection of the method will be achieved when it is possible to 



USE OF PUNCHED CARDS IN LINGUISTIC ANALYSIS 


371 


have cards that carry three lines of context punched on the same card. The 
amount of context carried in a line of 80 columns is not sufficient in a 
printed concordance prepared automatically. For the most part it is neces¬ 
sary to have a context of three lines, so that the word in question is always 
in the middle line. As already described, it is possible to obtain such print- 


r_ Deod Sea 
( Scroll card s 
In Scroll number 
sequence I 


Dead Sea Sc*ot 
words in inve/wT" 
/ / card to\on Magnetic Taf» 

(three larr^uage^ 


Words are inverted through 
Control Panel wiring 




Figure 16-6. Simplified block diagram of “Dead Sea Scrolls” processing on EDPM 
equipment. 




372 


PUNCHED CARDS 


ing even with standard machines, but repetitive use of the sorter and col¬ 
lator is necessary as well as extensive operations with the accounting 
machine. When such extensive operations can be avoided by using cards 
containing three lines of context, or some equivalent means, the mechaniza¬ 
tion of linguistic analysis can be said to have reached that stage where sub¬ 
stantial change will not be required for some time. 

When punched card systems operate so that the cards are passed through 
the machines along the cards’ long axis, w r hich is perpendicular to the axis 
of motion through present day machines with exception of the punchers 
and verifiers, then it would be possible to work on the basis of successive 
circuits, like a telephone center. The demands of linguistic analysis would 
then be satisfied even more completely, faster, and more economically. 

Such observations on desired developments should not overlook the fact 
that even with standard machines the punched card system permits more 
extensive, more certain, more advanced and more economical studies than 
would have been possible except with the patient work of many men. 

For similar work in the preparation of concordances by machine, the reader is 



Figure 16-7. Father Roberto Busa comparing the words of a modern scribe—the 
printing unit of an IBM 705 computer—with those written two thousand years ago 
by scribes of an ancient Hebrew sect living near the Dead Sea. 


USE OF PUNCHED CARDS IN LINGUISTIC ANALYSIS 


373 


referred to the work of Reverend James Ellison in preparing a concordance of the 
Bible by means of a computer. See, for example, 

“According to Mark 4.—,” Time, vol. 64, August 9,1954, pp. 68-9 
Soule, G. “Machine That Indexed the Bible,” Popular Science, vol. 169, Novem¬ 
ber 1956, pp. 173-5 

“Bible Labor of Years is Done in 400 Hours,” Lije, vol. 42, February 18, 1957, 
p. 92 

Cook, C. M. “Automation Comes to the Bible,” Christian Century, vol. 74, July 
24, 1957, pp. 892-4 


Appendix 

Work is now in process in Gallarate and in New York in applying the 
method of literary analysis described here to the task of cataloging the 
“Dead Sea Scrolls.” The nearly thirty thousand words under study were 
punched into IBM cards. A card was punched for each word, indicating 
its exact location and distinguishing characteristics. The entire set of cards 
was converted to two reels of magnetic tape by the IBM 705 computer in 
approximately two hours. See Figures 16-6 and 16-7. 

In order for the IBM 705 to perform the indexing of the “Dead Sea 
Scrolls,” the following items had to be taken into consideration. 

1. Card input requirements. 

2. Printed output requirements. 

Since Hebrew words are right-most justified, read and interpreted, special 
considerations had to be dealt with prior to obtaining the desired results. 
For the input, the Hebrew word cards are initially converted to tape in 
such a fashion that the words will be recorded on the magnetic tape in an 
inverted form (left-most justified). 

The Hebrew words range from 1 to 12 character positions. 

The magnetic tapes, once created, are then loaded on the IBM 705 and 
with the aid of modified Sort 57 Program Deck, the following has been ac¬ 
complished. 

1. This program first sorts all these Hebrew words alphabetically and at 
the same time re-inverts them into their original form prior to writing them 
on the output tape. 

2. An extension to the program provides for creating a summary word 
tape on which are written only those words which are graphically different 
from each other. This summary record will also show an identification serial 
number with the frequency of usage of each word. 

Later, on an off-line basis, these tapes will be listed on a tape-to-printer 
operation. The total off-line printing time is five hours. 



Chapter 17 

AN ABSTRACTING AND INFORMATION 
SERVICE FOR PLANT BREEDING 
AND GENETICS 


R. H. Richens 

Commonwealth Bureau of Plant Breeding and Genetics, School of Agriculture 

Cambridge, England 

A coding system in which all published articles on plant breeding and 
genetics are coded on punched cards for future reference is in use at the 
Commonwealth Bureau of Plant Breeding and Genetics, Cambridge. Since 
the principles involved appear to be of general interest for scientific study, 
an account of the technique used is being given here, in the hope that it 
may prove useful to others working in this field. 

General Principles of Mechanized Coding 

Mechanized coding of scientific papers involves the following processes: 
(1) Construction of a coding dictionary giving the equivalent in the code 
of any feature of a scientific paper. (2) Selection of those features of the 
paper that are to be coded. (3) Assigning to each of the chosen features its 
equivalent in the code. (4) Rendering the code into a medium susceptible 
of mechanized manipulation. It is necessary to consider each of these proc¬ 
esses in further detail. 

The construction of a coding dictionary must obviously precede the set¬ 
ting up of a mechanized coding system. A coding dictionary consists es¬ 
sentially of a series of entries giving for each possible feature of a scientific 
paper a unique equivalent in the code. The number of ways in which this 
might be accomplished is obviously very large. 

The simplest system would be to reproduce the feature to be coded with¬ 
out alteration. Thus, it might be agreed to code the word “wheat” when 
it occurs in a scientific paper by reproducing the word “wheat” in the code. 
Although this method has the advantage of extreme simplicity, subsequent 
mechanized manipulation is extremely intricate. The procedure could be 
refined by translating all linguistic equivalents of “wheat,” such as “bl6,” 
“Weizen,” “trigo,” “grano,” or “pszenica,” into one particular language— 
English, and it would be possible to go some way toward eliminating syn¬ 
onymy within a language, or within a scientific jargon. It could be decided, 
for instance, that “Triticum vulgare" should be used instead of “wheat,” or 
vice versa. 


374 



SERVICE FOR PLANT BREEDING AND GENETICS 


375 


Brevity is advantageous in most coding systems, especially when these 
are to be mechanized. Therefore, if it is possible to replace an expression, 
such as “Triticum vulgare,” by a unique equivalent with fewer letters, the 
efficiency of the coding system would probably be increased. There is no 
reason why the coding equivalent should be a succession of letters at all. 
It could be replaced by a series of numerals, a mixture of numerals and 
letters, a Chinese character, or some hieroglyph with no previously accepted 
significance. Convenience, however, and ease of recognition demand that 
the symbols used should be familiar, and it is, therefore, hardly necessary 
to consider coding equivalents made up otherwise than by the juxtaposition 
of letters and numerals. 

So far consideration has been given to a very simple type of coding, that 
in which single words occurring in a paper are coded. In most cases, how¬ 
ever, it is not the vocabulary of a paper but the ideas expressed that are 
coded. It is obvious that an idea cannot be reproduced as such in a code 
since ideas are characteristics of minds. All that can be done is to devise a 
code giving unique equivalents to each idea; these may be words or symbols 
as mentioned above. 

If ideas are being considered, a further very important refinement can be 
introduced into the coding system. Ideas exhibit manifold logical relations, 
one with the other. These relations, or some of them, can be reproduced in 
the structure of the coding equivalents. Thus, the idea of plant includes 
the notion of wheat, or, in logical terminology, wheat is a subclass of plants. 
Everything that is a member of the class of wheats is a member of the class 
of plants. Class inclusion occurs prominently in a system such as the Uni¬ 
versal Decimal Classification (U.D.C.), in which the coding equivalent for 
wheat, 633.11, is so constructed that it indicates that the wheat class 633.11 
is a subclass of cereals 633.1, and the cereal class, in turn, is a subclass of 
crop plants 633. 

The U.D.C. is only one, and, from a logical point of view, a somewhat 
unsatisfactory coding system. Alternative and better systems could be 
devised. Not only could additional logical relations between concepts be 
indicated in the structure of the coding equivalents, but a more complex 
and comprehensive system of class-inclusive relations could be used. The 
particular merits and demerits of the U.D.C. will be considered in the 
following. 

The second process in mechanized coding is the selection of those features 
of the original paper that are to be coded. 

The simplest case would again be constituted by vocabulary analysis. 
By selecting all the most scientifically significant words in a paper, a fairly 
adequate and detailed indication would be given of its contents. Moreover, 
this operation is in itself almost entirely mechanical, and requires no more 
than an ability to pick out significant words. If the procedure were reversed 



376 


PUNCHED CARDS 


so that all the words in a scientific paper in common speech—its “basic” 
vocabulary—were eliminated, and the residue taken as the significant vo¬ 
cabulary, the whole coding operation could be performed by a suitably 
constructed machine. 

To take an example, the scientifically significant vocabulary of a paper 
by Amason, Cumming, and Spinks is as follows 1 


aberrant 

microcurie 

aberration 

monococcum 

acetocarmine 

mutation 

aestivum 

Pelissier 

anaphase 

phosphate 

anther 

phosphorus 

beta 

pollen 

chromosome 

radioactive 

diaphane 

radium 

dutichon 

spikelet 

durum 

telophase 

einkorn 

tetraploid 

gamma 

Thatcher 

Hannchen 

translocation 

hexaploid 

Triticum 

Hordeum 

vulgare 

inversion 

localization 

X-ray 


These words indicate quite clearly the nature of the contents of the original 
paper, which is entitled “Chromosome Breakage in Plants Induced by Ra¬ 
dioactive Phosphorus (P 32 )”. 

Usually, however, the aspects of a paper to be coded are decided by a 
reader able to understand what the paper is about. The coder classifies the 
paper under certain general heads. He fits the paper into its place in a gen¬ 
eral classification of knowledge, however ill-defined the latter may be. For 
instance, in the case of the paper by Amason, Cumming, and Spinks this 
could be classified under three general heads: anomalous nuclear changes, 
botanical effects brought about by chemical agents, and radioactivity. In 
this classification three broad classes of phenomena have been selected, 
of which the principal phenomena described in the paper are representative 
members. 

It may seem that coding by vocabulary analysis and coding by meaning 
differ fundamentally, but it is possible to exaggerate this difference. In 
theory, at least, it would be possible to devise a mechanical procedure based 
on an analysis of the vocabulary and syntax of scientific papers, in which 
coding equivalents corresponding to the general classificatory heads just 
mentioned were assigned to the paper whenever certain combinations of 

1 Amason, Cumming, and Spinks, Science , 107 t 198-99. 



SERVICE FOR PLANT BREEDING AND GENETICS 


377 


words occurred, or even when certain words appeared in certain syntactical 
relations. Thus, it might be arranged that a paper containing any three of 
the following words, aberrant, aberration, deletion, interchange, inversion, 
nucleus, and translocation, should be given a coding equivalent correspond¬ 
ing to the general head “anomalous nuclear changes”, and similarly for 
other groups of words. This means that a purely mechanical analysis of 
vocabulary might be devised which would have the same result as a clas¬ 
sification under general heads assigned by an understanding reader. 

Attention has been given the seemingly impractical method of vocabulary 
analysis for coding scientific papers because it has theoretical interest and 
because it shows that it is possible to obtain mechanically results normally 
achieved only by making use of an understanding reader. This point may 
be put more precisely by stating that vocabulary analysis and semantic 
analysis of a scientific paper may produce isomorphic classifications. As 
far as practical problems go, however, the extraordinary efficiency of se¬ 
mantic analysis performed by the mind, as compared with the cumbrousness 
of purely mechanical methods, suggests that the latter are unlikely to be 
of practical importance in any small-scale scientific information center. 

Having now considered both the construction of coding dictionaries and 
possible ways of selecting the features of a paper that might be usefully 
coded, there is little to say on the third coding operation mentioned above, 
namely, assigning to each of the selected features of a paper its equivalent 
in the code. It is possible, since assigning coding equivalents is the setting 
down of a series of one-one equivalences, to construct a machine to do it; 
but here again a mechanical model is only likely to prove more efficient than 
the mind when the scale of the operation becomes very large. For compara¬ 
tively small-scale projects a human operator, using a check list and his 
memory, is likely to provide the most efficient means of assigning coding 
equivalents to the selected features of a scientific paper. 

The last process in mechanized coding, rendering the code into a medium 
susceptible to mechanical manipulation, has many possibilities. It is in some 
ways unfortunate that handwriting and print cannot be manipulated mech¬ 
anically without conversion, or translation in the logical sense, into a differ¬ 
ent medium. It is probable that the irregularities in handwriting and type 
setting are too great to enable any but the most complicated mechanical 
models to deal with their contents as efficiently as a human operator. It 
therefore seems necessary at the moment to render codes into appropriate 
functional media by employing a human intermediary. 

It is possible that the same individual may select the features of a paper 
to be coded, assign the coding equivalents, and render the latter into a func¬ 
tional medium. This is not necessary, however, and each of the processes 
could be performed by a distinct individual. Only considerations of effi- 



378 


PUNCHED CARDS 


ciency can decide how the human material in a coding system is to be dis¬ 
posed. 

Coding media present many possibilities, only a few of which have im¬ 
mediate practical interest. These are punched cards, photo-electrically de¬ 
tected light signals and electromagnetically sensed magnetic signals. The 
two latter media require apparatus beyond the reach of a small installation. 
They offer immense possibilities for very large-scale scientific information 
centers, but for small installations punched cards alone seem to offer an 
economical method. In this chapter, which is only concerned with coding 
methods suitable for small installations, attention will be confined to 
punched-card techniques. 

Outline of the System under Trial at the Commonwealth Bureau of 
Plant Breeding and Genetics 

(1) The coding dictionary used is the U.D.C. 

(2) Selection of the features of scientific papers to be coded is undertaken by the 
scientific staff of the Bureau. 

(3) The person who decides what aspects of a paper should be coded also assigns the 
coding equivalents, that is, a series of U.D.C. numbers. 

(4) The U.D.C. numbers are punched into cards by the clerical staff of the Bureau. 

Merits and Demerits of the U.D.C. for Mechanized Coding 

It has already been mentioned that the U.D.C. is a noteworthy example 
of a coding system whose structure is logically significant. There are several 
considerable advantages attached to its use. These are as follows: 

(1) The U.D.C. is very comprehensive, and a large proportion of the 
features of scientific papers that will require coding are to be found in it. 
Most organizations dealing with scientific information are interested pri¬ 
marily in the scientific aspect of their work, and not in logical principles of 
classification. For them, a ready-made classification, even though unsatis¬ 
factory on several scores, is likely to be of great service. 

(2) The U.D.C. is extensible, and its subject headings may be subdivided 
indefinitely. 

(3) The U.D.C. is internationally recognized. 

(4) The way in which U.D.C. numbers are constructed lends itself well 
to punched-card techniques. 

Against these advantages, there are three disadvantages: 

(1) It is not possible to draw together subjects which are placed far from 
each other in the U.D.C., and to subdivide the logical sum of the two classes, 
cytology + genetics for example, as a single unit. This is a particular dis¬ 
advantage in regard to the marked syncretic tendency of scientific studies. 
Subjects which once appeared remote from each other, as cytology, genetics, 
virus research, cancer research, the theory of evolution, paleontology, and 
ethnobiology, have now converged to such a degree that a discovery in any 



SERVICE FOR PLANT BREEDING AND GENETICS 


379 


particular one is quite likely to be of significance to one or more of the 
others. 

(2) The same subject is liable to appear in more than one place in the 
U.D.C. schedules. Thus, heredity is coded as 575.1 and 581.169, while 
hybridization is entered under 575.12 and 631.523. For purposes of mech¬ 
anized coding, those double entries have to be expurgated, and a choice 
made as to which alternative number in each case is to be used. 

(3) In some cases in the U.D.C. the principle of indicating the subclasses 
of a subject heading by suffixing further digits to it is not adopted. Thus, 
although basidiomycetes are given the number 632.44, the hemibasidiomy- 
cetes and eubasidiomycetes receive the number 632.45 and 632.47, respec¬ 
tively, instead of suffixing digits to 632.44. 

Apparatus 

“Powers-Samas” cards with 65 columns are used. Ten different colors are 
employed. The cards are punched by means of an ordinary “Powers-Samas” 
hand punch. The remaining item of equipment is a “Powers-Samas” sorter, 
having a selective sorting attachment. 


Coding Procedure 

In the Commonwealth Bureau of Plant Breeding and Genetics all ob¬ 
tainable scientific publications of possible interest to plant breeders and 
geneticists are read by the members of its scientific staff, who prepare for 
each paper an abstract in English for the Bureau publication, Plant Breeding 
Abstracts. In addition to summarizing each paper the abstracter also gives 
to it a U.D.C. code number as a guide to the nature of its contents. Papers 
that are only of indirect interest to plant breeder and geneticists are coded 
but not abstracted. 

The method of using the U.D.C. for coding purposes is the customary 
procedure of assigning a series of decimal numbers, each of which cor¬ 
responds to one aspect of the paper being coded. Thus, referring back to 
the paper of Amason, Cumming, and Spinks, already mentioned, this might 
be given the U.D.C. decimal numbers: 


576.356 

581.04 

539.16 


corresponding 
respectively to 
the three general 
heads 


anomalous nuclear 
changes 

botanical effects brought 
about by chemical agents 
radioactivity 


The coding numbers assigned by the Bureau are not exhaustive or even 
general descriptions of the contents of the papers coded. Papers are clas¬ 
sified from a particular point of view, that of its readers, and it is possible 
that their interest might be confined to a chance reference to plant breeding 



380 


PUNCHED CARDS 


in a paper covering principally some other topic. In this case, the U.D.C. 
numbers assigned by the Bureau will refer only to the relevant paragraph. 
A scientific information organization dealing with scientific papers in gen¬ 
eral or from a different point of view would code such a paper quite differ¬ 
ently. 

Layout of Card 

It has been found by experience that the subject matter of papers on 
plant breeding and genetics can be satisfactorily coded in most cases by one 
or more decimal numbers beginning with the following two digit combi¬ 
nations: 


51 mathematics 

53 physics 

57 biology 

58 botany 

63 agriculture 

In the layout of the card each of these subjects is assigned a separate 
panel of five columns (see Figure 17-la), while for the last item, agriculture, 
under which several subordinate aspects require separation, three five- 
column panels are used: 

631 agronomy 

632 phytopathology 

633-5 crops 

Should any U.D.C. numbers other than the above need to be employed, 
they are allocated to a separate five-column panel, entitled “General”. 
There are also certain suffixing decimal numbers which may be appended 
to any of the principal decimal numbers, such as .01 for bibliography. Such 
numbers are allocated to a further five-column panel, entitled “Tags”. 

All the panels of the card concerned with subject matter are of the same 
width—five columns. As far as the requirements of the coding scheme are 
concerned, this number of columns per panel is unnecessarily large in some 
cases. For instance, for the purposes of coding papers on genetics and plant 
breeding, the 51 and 53 panels dealing respectively with mathematics and 
physics could have been reduced. It has been found, however, that the 
work of punching is rendered easier if each of the subject panels is of equal 
width. In determining the precise layout of the card psychological consid¬ 
erations must not be disregarded. 

The remainder of the card is occupied by a thiee column panel for repre¬ 
senting the decimal number of the country to which a paper refeis, a three- 
column panel for representing its year of publication, a two-column panel 
for indicating the number of the volume of Plant Breeding Abstracts, in 



SERVICE FOR PLANT BREEDING AND GENETICS 


381 


which a summary of the paper has appeared, and a final one-column panel, 
entitled “Class,” in which a hole is punched in position 1 to indicate that 
the card is a subject card, thereby differentiating it from the other series of 
punched cards maintained by the Bureau. In the case of the three-column 
panel for the year only the three final digits of the year, for example 948 
for 1948, are punched. This will suffice for the next thousand years. Sub¬ 
sequent years can be differentiated by punching additional holes in the 
positions above 0. 

It is important to realize that the layout of the card is one of the most 
weighty factors determining the successful application of punched-card 
methods to scientific information work. A single mistake in layout will 
prejudice the whole of subsequent work. In general, an organization dealing 
with a comparatively narrow and highly specialized field will need a small 
number of wide panels. Those dealing with a wider range of subjects should 
use a larger number of narrower panels. Care should be taken that the 
frequency with which two decimal numbers are liable to fall within the same 
panel is low. It is also advantageous to have the panels of equal widths, 
even if it wastes space, since, as already mentioned, this lightens the work 
of the punching clerks. 

Preparation of Punched Cards 

After a paper has been coded by the scientific staff of the Bureau, the 
U.D.C. numbers are sent in to the clerical staff, together with the original 
paper to which they refer. The punching clerk then punches each of the 
digits of the U.D.C. numbers into the appropriate panels of the punched 
card by means of a hand punch, punching also the year of publication of 



Figure 17-1. Stages in the preparation of subject and author cards, a. Subject card 
before punching, b. Card after punching, c. Card after typing, d. An author card. 
















382 


PUNCHED CARDS 






Figure 17-1 (c) 



CUNMM, t «ttf 
•PINKS, J.W.T. 

87**M:MlO4:0Stj6 

(L) 

Chrooof at bmtefi In plants 

NllHiill Wf rVUMvTVVV 

nsotpnorut (p**i 

Soltfiec IS4St 107: !••-•• 

XVIII (4) 

Pmmmf. P. 


- • 


Figure 17-1 (d) 


the paper, the volume number of “Plant Breeding Abstracts,” in which the 
paper is noticed, and a hole in position 1 of the “Class” column to indicate 
that the card is a subject card. Figure 17-16 shows the mode of punching 
the coding data of the paper by Amason, Cumming, and Spinks mentioned 
























SERVICE FOR PLANT BREEDING AND GENETICS 


383 


previously. The first two digits of decimals coming in any of the first seven 
panels are not punched. The heading of the panel in these cases provides 
sufficient indication. 

After punching, the subject card is put in a typewriter above a standard 
filing card, with a carbon in between. The following information is then 
typed in duplicate on both cards: 

(1) The name of the author or authors. 

(2) The title of the paper, together with a translation if the original is not in English. 

(3) The other bibliographical details of the paper. 

(4) The volume of “Plant Breeding Abstracts,” in which the article is noticed. 

(5) The initials of the abstracter. 

(6) A note as to the library in which the original paper is deposited. 

(7) The coding decimals of the paper. 

The preparation of the subject card is now complete (Figure 17-lc). The 
standard filing card (Figure 17-ld) is used as an author card. 

The subject cards used are of ten different colors for convenience in 
filing. The color of the card is determined by the principal decimal number 
of the card, that is, the number recording the particular crop being treated. 
If there is no entry in the 633-5 crop panel in the coding details of the paper, 
that is, if the paper is a general one not dealing with any specific crop plant, 
uncolored Manila cards are used. In all other cases the color is determined 
by the final digit of the crop decimal, according to the following scheme: 

Final Digit Color 

1 Red 

2 Orange 

3 Yellow 

4 Green 

5 Blue 

6 Mauve 

7 Grey 

8 Pink 

9 Brown 

Thus, any paper on wheat, which will receive as one of its coding decimals, 
the number 633.11, will be entered on a red card since the final digit of 
633.11 is 1. Similarly, papers on potatoes, 633.491, will also receive red 
cards, while papers on barley, 633.16, will be given a mauve card. 

Should a publication deal with a number of unconnected subjects, as 
for example, the annual report of an agricultural research station, a series 
of cards will be made, one for each of the subjects included. Sometimes the 
coding data of a paper on a single topic contain two or more decimals that 
would come under the same panel, an eventuality which the card is designed 
to avoid as far as possible. When this happens, one decimal is punched in 
its proper panel, and the second (together with the first two digits normally 



384 


PUNCHED CARDS 


unpunched) is punched in the “General” panel. A second card is then 
punched in which these two numbers are reversed. Thus, a paper by Well- 
ensiek on “Methods for Producing Trilicales ”* was assigned the five deci¬ 
mal numbers: 


633.11 

575.129 

633.14 

581.04 

(49.2) . 


corresponding 
respectively to 
the subject heads 


wheat 

true breeding hybrids 
rye 

botanical effects brought 
about by chemical agents 
Holland 


For this paper two cards were prepared, the first (Figure 17-2a), a red card 
with the U.D.C. number 633.11 punched in its normal panel and 633.14 
punched in the “General” panel, and a green card (Figure 17-26) with 
633.14 punched in its normal panel and 633.11 punched in the “Gen¬ 
eral” panel. In both cases the suffixing decimal number .04, is punched in 
the panel headed “Tags,” as already explained. 

When two or more authors collaborate in writing a paper, a separate au¬ 
thor card is made for each. 


Filing 

The author cards are filed in alphabetical order in the usual way. 

The punched subject cards are filed according to crop since experience 
has shown that plant breeders’ inquiries normally relate to specific crops. 
Since the subject cards are colored according to crop, the cards will also 
be filed in color order. Within each color group corresponding to one U.D.C. 
number in the 633-5 crop panel the cards are not kept in predetermined 
order. Thus, all the subject cards relating to wheat, 633.11, are not filed 
in any particular order. Since all the cards pertaining to a particular crop 
are of the same color and the colors representing adjacent crops are dis¬ 
tinguished according to a standard plan, all that a filing clerk has to do when 
inserting a fresh card is to locate the approximate position of the crop, and 
insert the card anywhere within the run of cards of the same color already 
filed. 

Extraction of Information 

Requests for scientific information come in under various forms. Any 
inquiry relating to the published works of a particular author will be an¬ 
swered from the author file. All other inquiries are dealt w ? ith by passing 
appropriate subject cards through the sorter, and sorting either by a single 
column or by ten adjacent columns simultaneously by means of the selec¬ 
tive-sorting attachment. 

* Wellensiek, “Methods for Producing TriticalesJournal of Heredity, 38, 167-73. 



SERVICE FOR PLANT BREEDING AND GENETICS 


385 




Figure 17-2. Duplicate subject cards of a paper to which two decimal numbers have 
been assigned which would normally be punched in the same panel, a. A red card 
with 633.11 punched in its normal place and 633.14 punched in the “General” panel. 
b. A green card with these two decimals reversed. 

If, for instance, an inquirer should require a classified list of papers on 
cereal diseases, all the cards bearing the U.D.C. numbers 633.11 to 633.19 
would be extracted from the file and passed through the sorter set on col¬ 
umn 27, the second column of the 632 phytopathology panel. By this op¬ 
eration eight batches of cards would be segregated, dealing respectively 
with physiological diseases, galls, bacteria, fungi, angiosperm parasites and 
weeds, destructive animals other than insects, insects and viruses. 

Should any particular category in a one-column sort be of no interest, it 
can be ignored in the sort by suitable adjustment of the sorter. 

More frequently, an inquiry will take a more specific form, and may, for 




















386 


PUNCHED CARDS 


example, request details of all publications dealing with the resistance of 
wheat varieties to low temperature. To answer this all the wheat cards 
which bear the decimal 633.11 would be extracted from the file; their red 
color makes them easily discernible. The selective-sorting attachment is 
now fitted on to the sorter and set for the numbers 152162111 on columns 
21 to 29, respectively. By passing the wheat cards through the sorter it 
will now pick out the cards relating to all papers in which resistance to un¬ 
favorable conditions is mentioned, these having been given the decimal 
number (63) 1.521.6 beginning at column 21, the first column of the 631 
agronomy panel. However, these cards will be extracted only if they deal 
at the same time with low temperature, which is recorded by the decimal 
number (63) 2.111 beginning at column 26, the first column of the 632 phy¬ 
topathology panel. The panels of the cards are arranged so that subjects 
likely to occur in combination lie near one another. The extracting of the 
wheat cards from the file and the single passage through the sorter will now 
have provided information on all articles dealing with resistance to low 
temperature in wheat. 

Should the inquiry take an even more specific form, for instance, a request 
for all information on the genetics of resistance to low temperature in wheat, 
a visual inspection of the cards extracted as above may suffice. Alterna¬ 
tively, a second passage of these cards through the sorter, with the selec¬ 
tive-sorting attachment reset for genetics (57) 5.11 on column 11, the first 
column of the 57 biology panel, will serve the same end. Whenever two 
passages of the cards are required for selective sorting, the less frequently 
treated subject should be sorted out first, as the number of residual cards 
for the second selective sort will then be smaller. This rule applies also to 
hand-sorted cards, as noted in Chapter 2. 

Inquiries relating to papers concerned with particular countries or pub¬ 
lished at a particular time or over a particular period may be dealt with 
by setting the selective-sorting attachment over the “Country” and “Year” 
panels. It is also possible to eliminate unwanted categories during selective 
sorting. Thus, should an inquirer ask for all the papers dealing with rye that 
had appeared between 1940 and 1948, but excluding those that concerned 
North America, a single passage of the rye cards through the sorter, with 
the selective sorter set for 1940-48, but excluding cards bearing a hole in 
position 7, the U.D.C. number for North America, would give the required 
information. 

Standing Inquiries 

A scientific information unit may be required to supply its clients per¬ 
sonally with information of particular interest to them at regular intervals. 
This can be done mechanically by preparing one or more pilot cards for 



SERVICE FOR PLANT BREEDING AND GENETICS 


387 


each client and punching these with the coding equivalents of the subjects 
in which interest is expressed. Ever so often, the entire set of pilot cards 
is meshed with the subject cards acquired since the last similar operation. 
In this way, each pilot card comes at the head of a run of subject cards 
carrying the same punching and therefore dealing with the same subject 
as that referred to on the pilot card. The bibliographical details given in 
the run of subject cards can then be copied out and sent to the client or 
clients entered on the pilot card. 

Subject Index 

The annual subject index of an abstract journal is usually long and ardu¬ 
ous to compile. The Bureau has mechanized part of the process. The subject 
index takes the form of primary heading, subdivisions and cross references. 
Each coding equivalent has a corresponding index entry, which can be used 
either as a primary heading or subdivision. It also has, in most cases, a series 
of cross references attached to it. The subdivisions of the primary headings 
corresponding to any one panel of the subject card are arranged so that all 
are in one other single panel or, at least, in no more than three other panels. 

In making the index, all the subject cards of the year concerned are ex¬ 
tracted and sorted first into numerical order of one of the panels acting as 
subdivisor, and then into numerical order in the panel of the corresponding 
primary headings. All the cards relating to one particular set of primary 
headings and subdivisions are now assembled. A slip can now be made for 
each primary heading plus subdivision and the references to it can be 
transferred by hand from the relevant subject cards. This can be expedited 
in some cases by using accessory punched cards, punched so as to be sortable 
at the end of the whole operation into alphabetic order. Cross references 
can be made at this stage. 

The operation is then repeated for every pair of panels corresponding to 
a combination of primary headings and subdivisions. The slips can then be 
hand-sorted, or if accessory punched cards have been used, these can be 
machine-sorted, to give the final subject index, complete with primary 
headings, subdivisions and cross references. 




Part III 


FUNDAMENTAL CONSIDERATIONS 
IN CODING AND SYSTEMS 
DESIGN 




Part III 


FUNDAMENTAL CONSIDERATIONS 
IN CODING AND SYSTEMS 
DESIGN 




CHAPTER 18 

SUBJECT MAHER ANALYSIS AND 
CODING-SOME FUNDAMENTAL 
CONSIDERATIONS* 


James W. Perry 

Center for Documentation and Communication Research 
Western Reserve University, Cleveland, Ohio 


Introduction 

Valuable information often accumulates in amounts that require us to 
use aids to human memory. To serve their purpose efficiently, such aids 
must be organized in an orderly fashion. The development of such aids is 
a very old problem. The ancient Babylonians had classification systems for 
their libraries of clay tablets. Alphabetized indexing is at least as old as 
Gutenberg’s invention of printing. In recent years, the application of new 
tools, e.g., punched cards and electronic machines, have enabled new meth¬ 
ods to be developed for establishing and using orderly organization of in¬ 
formation. To develop new methods in the different forms that provide 
optimum advantages in dealing with different situations and circumstances, 
it is essential that the age-old problem of orderly organization of informa¬ 
tion shall be investigated anew in the light of the unusual capabilities of 
the new tools that are now finding increasingly widespread application. 

The organization of information for subsequent use is based on one form 
or another of the analysis of the information as to ideas, concepts, notions, 
abstractions, relationships, and the like. The procedures for conducting 
such analysis and, more particularly, the form of recording its results vary 
widely depending on the tools that we may choose to employ. For example, 
if we decide to designate various characteristics of the subject contents of 
documents by words and phrases and arrange them in alphabetical order, 
we have an index. If our analysis of documents as to characteristics is used 
to arrange the documents in groups and subgroups according to similarities 
and differences, we will establish a classification. 

These conventional methods of arrangement, indexing and classification, 
have been developed and used for the orderly arrangement of things in 
space. These methods function by establishing positional locations, in 

* This research was supported in whole or in part by the United States Air Force 
under Contract No. AF i9(6$8)-S67 monitored by the AF Office of Scientific Re¬ 
search of the Air Research and Development Command. 


391 



392 


PUNCHED CARDS 


ordered arrays on sheets of paper, on shelves or in drawers, or in separate 
pigeon holes. Arrangements in three-dimensional space encoimter practical 
limitations. For example, the classified shelving of books cannot simul¬ 
taneously group them according to year of publication, language and the 
many diverse subjects to which they pertain. Alphabetized indexes would 
be excessively bulky and prohibitively costly if they attempted to list every 
combination of characteristics which analysis of the subject contents of 
documents may reveal as important. Punched cards and electronic machines 
enable us to surmount such limitations of three-dimensional space. These 
newer tools are essentially multiple dimensional in character by virtue of 
their ability to search out and to select information on the basis of new 
combinations of characteristics, that is to say, combinations not formulated 
or established at the time the information is analyzed. Thanks to their 
multi-dimensional character of operation, the newer tools, as exemplified 
by punched cards, enable us to center our attention on selecting and re¬ 
trieving information for use. We are thereby released from the limitations 
of systems which function by deciding where we are going to put documents 
within a fixed array of one type or another. 

It is, of course, possible to base the operation of punched cards, electronic 
selectors and similar devices on the analysis of information in the form of 
a conventional subject matter classification or a simple index consisting of 
single terms: But more sophisticated concepts of subject matter analysis 
are required to tak e full advantage of these mechanical and electrical devices. 

We must break out of our mental strait jacket which has been imposed 
by accustomed means of subject handling: one-dimensional rows of words; 
two-dimensional sheets of paper; three-dimensional shelves and pigeon 
holes. We must stop thinking about where to put things. With punched 
cards and machines, we can do the equivalent of putting one thing in several 
places at the same time, of putting one thing just anywhere and retrieving 
it at will, of putting several things together somewhere and retrieving some¬ 
thing different. 

The “things” just mentioned are ideas or concepts that characterize in¬ 
formation, the “somethings” are combinations of ideas or concepts and 
relationships and the “places” may be exemplified by holes and notches in 
pieces of cardboard. In more general terms, the “places” are meaningfully 
disposed shapes and discontinuities in pieces of matter, or energy conduct¬ 
ing, generating or modifying spots and areas, such as magnetized spots on 
wire or cylinders, electric current conducting marks, light conducting, reflect¬ 
ing or absorbing spots and shapes, radioactive or fluorescent areas. To 
make use of such “places,” we must develop a code which assigns a unique 
pattern of discontinuities to designating each idea or concept, i.e.jjto each 



SUBJECT MATTER ANALYSIS AND CODING 


393 


characteristic of .subje ct, matt er. The construction and manipulation of the | 
array of discontinuities is done by appropriately designed mechanical or 
electrical devices, described in other chapters. This chapter will treat sub¬ 
ject matter analysis and codjng for the selection of desired information and 
its concommitant correlation. 

Statement of Basic Operations 

Other chapters in this book, especially Chapter 3, have directed attention 
to a variety of devices and equipment that have been applied for searching, 
selecting and correlating recorded information. On the one hand, we have 
relatively simple devices such as hand-sorted punched cards and, at the 
other extreme, fully automatic electronic selectors and computers. In spite 
of obvious diversity in design, the various devices and machines have certain t 
general operational functions in common. An understanding of basic prin¬ 
ciples underlying such operational functions is essential to developing codes 
for achieving efficient application of a given device or machine. Such under¬ 
standing can also provide guidance in selecting appropriate equipment for a 
given set of requirements and circumstances. 

The descriptions and summaries of practical applications as presented in 
this book make it clear that the use of various types of devices and equip¬ 
ment always involves four basic operations: 

1. Analysis of the subject contents of graphic records. 

2. Recording the results of analysis in an appropriate searching medium, 
such as, hand-sorted punched cards, magnetic tape, etc. 

3. Analysis of information requirements in terms of operations to be 
performed. 

4. Performance of searching and selecting operations. 

The first two steps are preliminary and preparatory in nature. The neces¬ 
sity for performing these steps arises from the limitations of machines that 
are available at present or that could be constructed at reasonable cost by 
exploiting present-day technology. Developments in character-recognizing 
devices may provide, within a few years, machines that are capable of scan¬ 
ning printed material and detecting individual words, punctuation marks, 
etc. Such detection is, however, quite different from selection of those 
words and phrases that correspond to important aspects of subject matter, 
and that might conceivably form the basis for writing abstracts, preparing 
indexes or encoding for machine searching. Furthermore, particularly with 
scientific and technical subject matter, diagrams, maps, graphs, equations 
and formulas constitute and often present a major part of the subject mat¬ 
ter and its correlation with textual material for abstracting, indexing, and 
encoding purposes usually requires a high degree of understanding of the 



394 


PUNCHED CARDS 


subject matter. 1 Programming a machine so that it provides such under¬ 
standing appears quite impossible with available equipment—or equipment 
that can be anticipated in the foreseeable future. To make the most effec¬ 
tive use of the machines with which we can expect to be working during the 
years ahead, it is necessary to provide an interpretation of the subject 
contents of graphic records in a form amenable to selecting and correlating 
operations that are performed by automatic and semi-automatic devices. 
A system for achieving such interpretation is often referred to as a “code” 
and its importance warrants considerable further discussion. 

The third and fourth steps outlined above pertain to the application of 
various mechanical or electrical devices to collections of encoded informa¬ 
tion to identify, to select and to correlate items of pertinent interest. The 
third step)—the analysis of information requirements—is, in character at 
least, closely akin to the preliminary encoding of input information. Just 
as searching machines available at present require that input information 
shall be preliminarily analyzed and encoded, so also presently available de¬ 
vices and equipment require that questions to be answered and requests 
for information in general shall be submitted to preliminary analysis. Such 
analysis must be conducted on the same basis as the preliminary analysis 
and coding of the information to be searched. The results of such analysis 
of questions and information requests serve to guide the selecting and corre¬ 
lating operations to be performed by various mechanical and electrical de¬ 
vices. With hand-sorted punched cards, for example, the analysis of infor¬ 
mation requirements will lead to decisions to perform one or more sorting 
operations directed to various punching positions. When a plurality of such 
sorting operations are involved, the analysis of information requirements 
may suggest possibilities for conserving considerable time and effort in 
manual manipulation of the cards. With more elaborate equipment, particu¬ 
larly fully automatic electronic selectors, the analysis of information re¬ 
quirements provides instructions for conditioning the machine to perform 
desired selecting and correlating operations. Such conditioning is often re¬ 
ferred to as programming and, with some machines at least, may be ac¬ 
complished by appropriate wiring of plugboards. 

The fourth step—performance of searching and selecting operations— 
may be simple or complex in nature depending both on the information 

1 It should be noted in this connection that the automatic abstracting of news 
articles, as published in the weekly magazine Time, has been reported recently by 
H. P. Luhn, “The Automatic Creation of Literature Abstracts,” IBM Journal of 
Research and Development, 2, No. 2, 159-165 (April 1958). For a report of results of 
application of the same techniques to abstracting scientific review papers as pub¬ 
lished by the Sci. American, see H. P. Luhn, paper presented at the symposium on 
documentation, School of Library Science, University of Southern California, Los 
Angeles, April, 1958. 



SUBJECT MATTER ANALYSIS AND CODING 


395 


requirement and on the operational characteristics of the equipment being 
used. Even with relatively simple equipment, such as hand-sorted punched 
cards, it is instructive to formulate searching operations on a logical basis. 
(See the next chapter for details.) With fully automatic searching selectors, 
such formulation is a virtual necessity for efficient programming. For concise 
expression of such formulations for programming, the Boolean notation is 
convenient and easy to apply, as discussed subsequently in this chapter and 
elsewhere.* 

From what has been said above, it is perhaps obvious that the ability to 
perform searching, selecting and correlating operations is dependent on, 
and determined by, the preliminary analysis of input material, i.e., its en¬ 
coding. If a high degree of reliability in accomplishing the desired selecting 
and correlating operations is required, then corresponding care must be de¬ 
voted to achieving consistency during the analysis-encoding steps. Further¬ 
more, the higher the required degree of selectivity and of correlating ability 
the more care must be devoted to establishing the rules for analysis of in¬ 
formation and for encoding the results of such analysis. Selectivity is de¬ 
termined, not so much by the complexity of the system for analysis and 
encoding, but rather by the care with which such a system is designed. 
Each element of complexity in the system must justify itself by providing 
a useful contribution to advantageous selectivity. 

Investigation of the relationships between various forms of complexity, 
on the one hand, and useful selectivity, on the other hand, is still going 
forward and is still far from complete. At one time, not so long ago, it was 
necessary to rely almost entirely on judgment and hunch in establishing 
systems for analyzing and encoding information for mechanical and elec¬ 
trical selection. This situation has improved considerably during recent 
years, thanks both to the continuing efforts of a wide circle of workers and 
also to their willingness to share their experience as written up, for exam¬ 
ple, in the chapters of this book. By taking such experience into account, 
it is possible to avoid many pitfalls. It is the purpose of this chapter to 
summarize such experience and thus provide guide lines for exercising 
judgment in developing systems for analyzing and encoding information 
for searching by mechanical and electrical means. 

Coding—Its Nature and Role 

As already noted, the purpose of preliminary analysis and encoding of 
information is to make it possible to apply various mechanical and elec- 

* See, J. W. Perry, Allen Kent, and M. M. Berry, “Machine Literature Search¬ 
ing,” Chapter 6, New York, Interscience Publishers, Inc., 1956; Jessica Melton, 
and J. W. Perry “Analysis of Questions,” Chapter 15 in J. W. Perry and Allen Kent, 
“Tools for Machine Literature Searching,” New York, Interscience Publishers, 
Inc., 1958. 



396 


PUNCHED CARDS 


trical devices to search, to select and to correlate the information contained 
in a given collection of graphic records. To this end, the subject contents 
of graphic records are analyzed with respect to those features which can 
serve as the basis for performing searching, selecting and correlating opera¬ 
tions by means of various devices or machines. The results of such analysis 
must be recorded in a form that enables the searching, selecting and cor¬ 
relating equipment to respond to individual features of the subject contents 
of records and to combinations of such features. Depending on the equip¬ 
ment used for searching, selecting and correlating this recording assumes a 
variety of forms. With hand-sorted punched cards, notches are cut at pre¬ 
determined positions along the cards’ periphery. With “Uniterm” cards, 
document numbers are written or otherwise posted on cards each of which 
corresponds to some one subject heading (see Chapter 7). With magnetic 
tape, invisible patterns of magnetic spots are produced. With “Minicards”' 
photographic procedures are used to produce patterns of opaque and trans¬ 
parent spots. 

These examples of different modes of recording may suffice to make the 
point that the recording of features of subject matter is in a form whose 
purpose is to permit mechanical or electrical devices to perform selecting 
and correlating operations. This form of recording is quite different from 
printing, writing, diagramming, etc., to which we are accustomed. Recorded 
material for mechanical or electrical searching is, accordingly, often spoken 
| of as “coded.” The term “code” is often applied to rules for generating re- 
I cordings for automatic or semi-automatic searching. The more complex 
( codes that may be used to advantage with certain types of automatic equip¬ 
ment are sometimes termed “machine language.” 

Codes may range, therefore, from lists of words or terms for use with 
simple devices (e.g., hand-sorted punched cards, “Uniterm” cards) to 
relatively elaborate and carefully designed artificial languages, which re¬ 
semble machine Esperanto and into which statements in human languages 
tmay be interpreted. An essential part of a code is a set of rules for recording 
Iselected features of subject matter, e.g., by punching cards, or magnetizing 
/spots on tape, so that one device or another may be used for selecting and 
correlating operations. 

It is perhaps possible to accomplish encoding in such a way that it may 
appear to be a single step. With a simple device, such as hand-sorted 

* A. W. Tyler, W. L. Myers and J. W. Kuipers, “The Application of the Kodak 
Minicard Equipment to Problems of Documentation,” Am. Doc., 6, 18-30 (1955). 
cf. also: “Minicard Demonstration,” ibid, 258-259; “A Minicard System for Docu¬ 
mentary Information,” Chapter 27 in J. H. Shera, Allen Kent and J. W. Perry (eds.), 
“Information Systems in Documentation,” New York, Interscience Publishers, 
Inc., 1957. 



SUBJECT MATTER ANALYSIS AND CODING 


397 


punched cards, encoding is sometimes practiced by inspecting the informa¬ 
tion, e.g., reading an abstract typed or mounted on the card, and immedi¬ 
ately proceeding with the punching of the card. Even in such a case, certain 
punching positions must be assigned meaning and this assignment must be 
kept in mind while encoding. The possibilities of making mistakes and the 
desirability, not to say necessity, of consistency in encoding make it ad¬ 
visable, even with simple codes, to proceed in a more formal and deliberate 
fashion, to carry out the following sequences of steps, and consciously to 
distinguish between them: 

1. To decide what features of subject matter shall be expressed by coding. 

2. To record such decisions as to important features in a well-defined, 
orderly fashion. (Preparatory to coding hand-sorted punched cards, for 
example, an orderly list of important subject headings may be prepared. 
When applying a comprehensive “machine language,” it may be continually 
extended to include new terms and concepts as they are encountered in 
papers and reports undergoing encoding 4 .) 

3. To decide how each feature will be recorded for mechanical or elec¬ 
trical searching and selecting. (With hand-sorted punched cards, for ex¬ 
ample, this type of decision will involve assigning some one punching posi¬ 
tion or some combination of positions to record a given feature of subject 
matter as expressed, for example, by a corresponding subject heading.) 

In this way a set of rules or procedures are established and followed for 
accomplishing the analysis and encoding of information. 

It is, perhaps, instructive to point out certain similarities with subject 
indexing. Decisions must be made as to what features of subject matter 
are of sufficient importance to be expressed by coding or by index entries. 
It is essential both for good indexing and good coding that these decisions 
shall be made in a consistent fashion in reviewing the subject contents of 
successive documents. It is equally important that the results of decisions 
as to important features of the subject contents of documents shall be re¬ 
corded in an unambiguous and consistent fashion. In subject indexing this 
leads to the requirement that terminology shall be used in a consistent 
fashion in setting up index entries. To meet this requirement, lists of care¬ 
fully compiled, well-defined subject headings are often used when index¬ 
ing. The same purpose may be served, especially when coding hand-sorted 
punched cards, by lists of subject headings which also indicate how the 
cards are to be punched to record the individual items. (See, for example, 
Chapters 4, 5, 8,14). With fully automatic searching selectors, simple lists 

4 See J. W. Perry, Allen Kent and M. M. Berry, “Machine Literature Searching,” 
Chapter 11, New York, Interscience Publishers, Inc., 1956; Jessica Melton and 
J. W. Perry, Chapter 5, in J. W. Perry and Allen Kent, “Tools for Machine Lit¬ 
erature Searching,” New York, Interscience Publishers, Inc., 1958. 



398 


PUNCHED CARDS 


of terms are not the most effective means for accomplishing the analysis 
and encoding of recorded information. But with such selectors it remains 
true, of course, that important features of the subject contents of docu¬ 
ments must be recorded in an unambiguous and consistent fashion. To this 
end, systematic rules, whose aggregate is sometimes termed “machine 
language,” have been developed. 4 

The development of an appropriate code—in other words, a set of 
rules for analyzing subject matter and recording the results of such anal¬ 
ysis—is no simple matter. This statement remains true, even when the code 
is being developed for such relatively simple devices as hand-sorted punched 
cards. The cards—rather obviously—are incapable of thinking. Rather 
they are devices for performing certain operations which can be con¬ 
ducted so as to accomplish useful selections. Similarly, the most advanced 
searching selectors perform certain operations—albeit more complex in 
nature—that accomplish useful selections of a more versatile type. 

It is perhaps evident that code development must take into account the 
nature of such selecting operations that can be performed with the aid of 
various mechanical and electrical devices. 

The various devices with which this book is concerned permit, in one 
fashion or another, a number of characteristics of a given paper, patent, 
document or graphic record to be independently searched. As a consequence, 
searching and selecting operations may be directed to any one of the 
characteristics so recorded. This capability, of itself, is not sufficiently 
useful or interesting to insure widespread application of searching equip¬ 
ment no matter how flexible or rapid in operation.* The effectiveness of 
all kinds of searching and selecting equipment—from hand-sorted punched 
cards to fully automatic electronic selectors—results, rather, from their 
ability to perform searching operations that are defined in terms of com¬ 
binations of characteristics. Some simple considerations suffice to indicate 
the underlying principles. 

Selecting Operations and Equipment Capabilities 

As pointed out in various chapters in this book and elsewhere,® a con¬ 
siderable variety of mechanical and electronic devices have been applied 

* Ralph R. Shaw, “Machines and the Bibliographical Problems of the Twen¬ 
tieth Century,” in Louis N. Ridenour, Ralph R. Shaw and Albert G. Hill, “Bib¬ 
liography in an Age of Science,” Urbana, University of Illinois Press, 1952; Carl S. 
Wise and J. W. Perry, “Multiple Coding and the Rapid Selector,” Am. Doc., 1, 76-83 
(1950). 

• Marjorie R. Hyslop, “Inventory of Methods and Devices for Analysis, Storage 
and Retrieval of Information,” Chapter 6 in J. H. Shera, Allen Kent and J. W. Perry 
(eds.), “Documentation in Action,” New York, Reinhold Publishing Corporation, 
1956. 



SUBJECT MATTER ANALYSIS AND CODING 


399 


to achieve advantageous selection and correlation of recorded information. 
To characterize their capabilities and limitations in an exhaustive fashion 
would require a lengthy treatise. In spite of obvious differences in mode of 
functioning, such diverse devices as hand-sorted punched cards and 
electronic selectors have a number of important operational characteristics 
in common. 

1. Independent recording of characteristics of the subject contents of 
documents. (The range and nature of characteristics that can be con¬ 
veniently and advantageously recorded does vary, however, depending on 
the device or equipment being used.) 

2. Searching and selecting operations may be directed to any one of 
the independently recorded characteristics or to their combinations. (The 
range of such combinations depends not only on the extent to which 
different types of characteristics may be recorded but also on the ability 
of various searching and selecting devices and equipment to detect various 
characteristics and to respond to them.) 

3. Concise statement of searching and selecting operations may be 
accomplished with the notation of Boolean algebra. (Chapter 19 in¬ 
dicates how the Boolean notation may be applied to specifying various 
selecting operations performed with hand-sorted punched cards. Similar 
formulations are also applicable to aspect cards, e.g., “Peek-a-boo,” “Uni¬ 
term,” and the like. Furthermore, the same mode of abstractly specifying 
selecting operations has also been used both in designing fully automatic 
equipment and in formulating procedures for encoding abstracts. 7 ) 

Subsequent discussion may serve not only to show how the performance 
characteristics of various kinds of equipment may be formulated, but also 
to indicate the wide range in capabilities and limitations of present-day 
devices. Selection of a given device or combination of devices to meet a 
given situation can be expected to require its careful analysis in terms of 
purposes to be served, the value of benefits that can be provided and the 
costs of conducting various operations, both intellectual and routine in 
nature. 8 

With punched cards, for example, individual holes may be punched to 

7 For a summary of underlying principles, see: J. W. Perry, Allen Kent and M. M. 
Berry, “Machine Literature Searching,” Chapters 11 and 13, New York, Intersci¬ 
ence Publishers, Inc., 1956; J. W. Perry and Allen Kent, “The New Look in Library 
Science,” Appl. Mechanics Revs., 9, No. 11, 457-60 (1956). See also, for fully detailed 
presentation: J. W. Perry and Allen Kent, “Tools for Machine Literature 
Searching,” New York, Interscience Publishers, Inc., 1958. 

* For an introductory discussion of a general theory for analysis of costs and 
benefits, see J. W. Perry and Allen Kent, “Documentation and Information Re¬ 
trieval,” New York, Interscience Publishers, Inc., 1957, cf., also Chapter 9 in J. W. 
Perry, Allen Kent and M. M. Berry, “Machine Literature Searching,” New York, 
Interscience Publishers, Inc., 1956. 



400 


PUNCHED CARDS 


designate various subject matter characteristics. Alternately various 
combinations of holes may be punched so that some combination denotes 
a certain characteristic as designated when establishing the code. A search 
for needed information may then be formulated in terms of various char¬ 
acteristic combinations each of which is recorded either by punching a 
single hole (direct coding) or by punching a combination of holes (more 
complex codes as exemplified by numerical codes, random superimposed 
codes, and the like). In discussing such selecting operations, and in for¬ 
mulating them concisely, it is convenient to use capital letters (A, B, C, D, 
etc.) to designate, with hand-sorted punched cards, a certain hole or 
combination of holes to which meaning has been assigned. More generally, 
capital letters, with subscripts, may be used to designate meaningful 
recorded patterns and their combinations at various levels corresponding, 
for example, to “words,” “phrases,” “sentences,” and the like, in machine 
language for use with fully automatic selectors as summarized previously in 
chapter 11. 

One of the simplest, and most important forms of combinations of 
characteristics may be expressed as the search requirement that all of 
several (two or more) characteristics shall be detected. With hand-sorted 
punched cards, for example, a search may be conducted by selecting out 
those cards that have all of several (two or more) positions (or meaningful 
combinations of positions) punched so as to respond to sorting operations. 
As discussed in detail in Chapter 19, such sorting operations may be 
designated symbolically as 

AB-CD - 

where the individual letters indicate the various positions that may be 
punched in each of the cards, so that those cards that are punched may 
be separated from others that are not so punched. With aspect cards, a 
similar search may be conducted by observing those document numbers 
that are present on all of several (two or more) of the aspect cards. 

Similarly, we may select out those cards that have been punched at 
any one, at least, of several punching positions. Then we may select those 
cards that have been punched at position A and/or position B, and/or 
position C, and/or position D, etc. Such a sorting operation is conveniently 
and concisely symbolized by the logical sum, an example of which fol¬ 
lows: 


A + B + C + D, etc. 

A third possibility is to reject those cards that are punched in a given 
position. For example, we may reject those cards punched in position B 
while accepting those punched in position A. Such a specification of search 
is an example of a logical difference and may be symbolized by A — B. 



SUBJECT MATTER ANALYSIS AND CODING 


401 


It is theoretically possible to carry out, with hand-sorted punched cards— 
see Chapter 19—(or aspect cards), more complex sorting (or matching) 
operations of any degree of complexity which involve two or more of the 
three above mentioned logically defined basic operations. Such complex 
sorting (or matching) operations may be exemplified as follows: 

(A + B) (C + D) 

(A-B) + (C-D) 

l(A-B) + (C-D)] [(E - F)] 

(E - F) + (C-D) 

As performed with hand-sorted punched cards, the logically defined 
sorting operations relate to the holes and notches cut in the cards. These 
operations, strictly speaking do no more than separate the cards that are 
characterized by certain holes and notches. With aspect cards, the logically 
defined number-matching operations do no more than identify certain docu¬ 
ments that are characterized by having their serial numbers or similar 
identification recorded on the aspect cards involved in a particular search. 

The practical limitations with regard to complexity of logically defined 
operations that can be performed conveniently show up in different forms 
and degrees, depending upon the type of searching device or principle used. 
With hand-sorted punched cards, the number of cards to be “needled” as 
the size of the file increases usually provides the first trials of patience in 
conducting manual operations. With aspect cards, the first trials of 
patience show up as the complexity of searching operations increase, partic¬ 
ularly with regard to the conducting of selecting and correlating procedures 
that involve logical products one or more of whose terms is a logical sum. 

Assigning meaning to the holes and notches of punched cards (or to the 
cards of the aspect system) is an entirely independent operation which is, 
of course, essential to purposeful use of the cards to accomplishing useful 
searches, selections and correlations. 

When meaning is attributed to each individual punched position, the 
form of coding is termed “direct coding.” Meaning may also be ascribed to 
each of various combinations of holes in a wide variety of ways. The two 
main types of such combinational codes employed with hand-sorted 
punched cards may be distinguished as follows: 

1. In a given field, (i.e., in a given group or set of holes) only one com¬ 
bination may be punched in any one card. (But, of course, within the same 
field, different combinations may be punched in different cards. Examples 
of this form of coding are numerical and alphabetical codes described in 
Chapter 2, pages 18-21.) 

2. In a given field, in any one card, more than one combination may 
be punched. (The number of such combinations that may be advan- 



402 


PUNCHED CARDS 


tageously punched in any one card will, however, be limited by the pos¬ 
sibility of undesired fortuitous generation of false or “ghost” combinations 
and by the accompanying possibility of cards being selected on the basis 
of “ghost” combinations that do not correspond to intended code entries 
for such cards 8 . To counteract the tendency to generate false or “ghost” 
combinations in this form of coding, random numbers are often used to 
establish each of the combinations to which meaning is assigned. From 
these practices, the name “random superimposed coding” has been de¬ 
rived.) 

Standard coding for machine manipulated cards calls for punching one 
or more holes in a column of a card to designate a single letter, numeral 
or other symbol. (See Chapter 3, pages 55, 65). Both the direct and 
random superimposed form of coding have also been applied to machine 
manipulated cards which may be sorted and selected by various machines 
including relatively low-cost sorters, as available, for example, from 
IBM or Remington-Rand. 

Coding for aspect systems using machine manipulated cards may involve 
punching document numbers, one to a card, on separate decks of cards, 
each deck representing a code, a word, or an idea. Identification of desired 
documents may be effected either by automatic routines for collating the 
decks of cards representing each of the words or ideas involved in a search, 
or by visually identifying number matches. 

The mechanical sorters used in both of these approaches are so designed 
and constructed that advantageous performance, in a practical sense, of a 
searching operation directed to a given hole (or combination of holes) re¬ 
quires that it be known in which column the hole (or combination of holes) 
in question has been punched. This remains true when a sequence of col¬ 
umns (or “field”) is used to record no more than a single meaningful com¬ 
bination of holes in any one card or when, by applying random superim¬ 
posed coding, a plurality of meaningful combinations are punched in a 
given “field” of a single card. With both these approaches to coding, the 
columns are organized into rigidly defined sets or “fixed fields” as they are 
usually termed. 

Coding based on fixed fields is subject to severe limitations, as was 
understood nearly a decade ago 10 . In particular, with a limited number of 
fixed fields, it is necessary, for reliable operation, to establish a corre¬ 
sponding organization of the subject characteristics of the information to 

• See Carl S. Wise, “Mathematical Analysis of Coding Systems,” Chapter 20 
(in particular pp. 285-299) in Robert S. Casey and J. W. Perry (eds.) “Punched 
Cards. Their Application in Science and Industry,” New York, Reinhold Publish¬ 
ing Corporation, 1951. Revised and reprinted as Chapter 21 of this book. 

10 J. W. Perry, “The ACS Punched Card Committee. An Interim Report,” Chem. 
Eng. News, 27, 754-756 (1949). cf., also ibid, 28, 3789 (1950). 



SUBJECT MATTER ANALYSIS AND CODING 


403 


be encoded. Similarly, for reliable as well as efficient operation of an aspect 
system, the requirement to devote a single physical record to an aspect repre¬ 
senting an area of subject matter makes it necessary to make arbitrary de¬ 
cisions for controlling the scope or range of subject matter covered by each 
aspect record. With certain kinds of information it is relatively easy to 
organize the important characteristics into mutually exclusive sets with 
one characteristic in each of the sets being both necessary and sufficient 
for a given item of information. If, for example, the item of information 
is a business transaction such as the selling of a certain shipment of some 
one product, then the various sets of characteristics will be typified by (i) 
the name, code number or other designation of the product (ii) the amount 
sold (iii) the unit price (iv) the customer to whom sold (v) the salesman 
involved (vi) date of sale (vii) date of delivery, and so on. One character¬ 
istic each from such sets provides the needed information regarding the 
sale, and assignment of appropriate fields on the punched cards enables 
the various characteristics to be recorded in the preassigned fixed field. 
Such sets of characteristics are spoken of as “mutually exclusive” for the 
reason that specification of any one characteristic within a given set 
excludes the possibility of another characteristic within the same set 
being pertinent to a given item of information. Thus, specification of one 
unit price for a given shipment of product excludes the possibility that 
another unit price should apply to the same shipment. The sale to one 
customer excludes simultaneous sale of the same shipment to another 
customer. The delivery on one date excludes delivery of the same ship¬ 
ment on another date, etc. It should also be noted that these sets of char¬ 
acteristics are also mutually independent. Thus the unit price of a product 
may change independently of the amount sold, or the customer may vary 
independently of the date of sale, etc. 

In many practical cases of characterizing recorded information some 
of the characteristics may be organized into mutually exclusive sets, while 
other characteristics are not amenable to such treatment. For example, 
personnel data relating to various individuals involves certain sets of 
characteristics that are mutually exclusive. Thus, birth at one date ex¬ 
cludes the possibility of birth on some other date. Similarly, birth at some 
one town or other location excludes birth somewhere else. Thus, birth 
date and place of birth constitute examples of sets of characteristics 
that are mutually exclusive. On the other hand, knowledge of one language 
does not exclude the possibility of knowledge of one or more other lan¬ 
guages. Similarly, attendance at one university or college does not exclude 
the possibility of attendance at another being included in personal data. 
Attempting to fit such characteristics as knowledge of languages and 
attendance at universities or colleges into a fixed field scheme is sure to lead 



404 


PUNCHED CARDS 


to difficulties. When the subject contents of documents pertain to a wide 
range of characteristics that cannot be organized into a restricted number 
of mutually exclusive sets, the fixed field system of machine-sorted punched 
cards presents difficulties which become more severe as the range of such 
characteristics becomes wider. Even with moderately broad ranges of 
characteristics the attendant difficulties impose limitations whose sur¬ 
mounting justifies considerable effort. 

As pointed out in Chapter 2 (pages 55, 65) fixed field coding with 
machine-sorted punched cards is usually carried out by recording a single 
symbol, especially some one numeral or letter, in a single column of the 
card. With IBM cards, for example, this means that the twelve punching 
positions in a column are being used to record a single symbol. Direct 
coding—i.e., assignment of a definite well-formulated meaning to each 
individual hole in a specified field in a machine-sorted punched card—is a 
simple form of coding which enables more than one entry to be made in 
one column. This method has been applied with impressive success in 
dealing with narrow ranges of subject matter 11 . But it is now widely 
recognized that this approach is limited in the range of subject matter 
that may be encoded in sufficient detail to provide discriminating and 
correlating capabilities such as are needed in dealing with extensive files 
of complex information as encountered, for example, in science and tech¬ 
nology. Another way in which the limitations of fixed fields may be partially 
surmounted is by the use of random superimposed coding. (See Chapter 
10.) With this method, a restricted number of characteristics within a 
given set may be recorded in a given card in the fixed field reserved for the 
set of characteristics. The restriction in number of such characteristics 
that may be recorded on a single card will depend on the number of punch¬ 
ing positions in the field, the number of holes in a meaningful combination, 
the extent to which false sorts may be tolerated and related factors. 

In aspect systems, the intellectual analog of the fixed field coding problem 
is encountered because of the operational requirement to organize charac¬ 
teristics into sets, in order to achieve practical results during the exploita¬ 
tion phase. Because of the practical limitations encountered in aspect sys¬ 
tems in performing searching and correlating operations involving logical 
products of series of logical sums, the need to reduce scatter of synonomous 
or partially synonomous terminology which may be used as headings of 

11 Julius Frome and Jacob Leibowitz, “A Punched-Card System for Searching 
Steroid Compounds.” Patent Office Research and Development Reports. No. 7, 
Washington, Dept, of Commerce, July 8, 1957. Don. D. Andrews. “Progress Report 
on U. S. Patent Office Mechanized Searching” Paper presented before the Division 
of Chemical Literature, American Chemical Society, 113th National Meeting, 
San Francisco, California, April, 1958. 



SUBJECT MATTER ANALYSIS AND CODING 


405 


aspect cards becomes controlling. This consideration underlies the coding 
described in Chapters 6 (pages 141-146) and 7. 

In order to go further in surmounting the limitations of fixed field 
coding so that broader ranges of subject matter could be rendered amenable 
to mechanical or electrical searching and selecting operations, various 
machines have been specially designed in recent years. The Luhn Scanner 
was designed to pass IBM cards end-wise and to detect various letters or 
other symbols as distinctive combinations of five holes in the twelve 
punched positions within any one column. The machine detected in¬ 
dividual symbols or meaningful sequences of symbols so recorded in a 
fashion analogous to a blind man reading Braille and detecting an in¬ 
dividual letter, e.g., “c”, or a letter sequence, e.g., “c—a—t.” The ability 
of the machine to detect combinations of meaningful symbols or their 
sequences as exemplified by words or codes was restricted, however, to a 
single level of logical combinations. 1 * 

Recent machine development work has surmounted this limitation. It 
has been found possible to construct—at a cost not exceeding a fraction 
of the price of a general-purpose computer—searching equipment 13 charac¬ 
terized by the following capabilities: 

1. Symbols and symbol sequences may be entered in the recording 
medium, e.g., magnetic tape, without limitations of the fixed field type. 
Symbols and symbol sequences are detected as code patterns of the same 
type as used with Teletype tape or the above mentioned Luhn Scanner. 

2. Symbols and symbol sequences may be organized into combinational 
levels analogous to “syllables,” “words,” “phrases,” “sentences,” “para¬ 
graphs,” etc. Letters with subscripts provide a simple notation for dis¬ 
tinguishing such levels, thus we have: 

Aj, Bi, Ci, Di, etc., for “syllables” 

A 2 ,B 2 ,C 2 ,D 2 , etc., for “words” 

12 Staff report, “Machine Techniques for Information Selection,” Chem. Eng. 
News, 30, 2806-10 (1952); H. P. Luhn, “The IBM Electronic Information Search¬ 
ing System,” International Business Machines Corp., Poughkeepsie, New York, 
1952. 

M J. W. Perry, “The Western Reserve University Searching Selector,” Chapter 
18 in J. W. Perry and Allen Kent “Tools for Machine Literature Searching,” New 
York, Interscience Publishers, Inc., 1958. Note also that recent studies and tests 
have demonstrated that appropriate programming enables at least certain elec¬ 
tronic computers, in particular the IBM 650 and the IBM 705, to accomplish 
searching and selecting operations as specified above. See Sally F. Dennis, “Pro¬ 
gramming the IBM 650 to Searching Encoded Abstracts,” Chapter 19 in J. W. Perry 
and Allen Kent “Tools for Machine Literature Searching,” New York, Interscience 
Publishers, Inc., 1958. 



406 


PUNCHED CARDS 


A 8 , B 3 , C*, Dj, etc., for “phrases” 

A 4 , B 4 , O 4 , D 4 , etc., for “sentences” 

A 6 , B 8 , C6, D 6 , etc., for “paragraphs” 

A«, B # , C B , D®, etc., for “messages” 

3. At any level, which we may term the “n-th” level, an encoded com¬ 
bination will consist, as a rule, of a number of component combinations 
at the “n — 1 ” level. Each of several “n-th” level combinations, denoted by 
A n , B n , C n , D„ , etc. may be specified in terms of component elements 
designated by A n -i, B„_i, D„_i, etc., or more generally in terms of lower 
level component elements exemplified by, A«, Bj, Ak, C m , D*, R f , 
Af, C e , etc., where j, k, m, e and f denote levels lower than n. As discussed 
in detail in Chapter 11 pages 252-256, combinations of elements may be 
specified at any level in terms of logical product, sum and difference and 
their complex combinations. 

The construction of moderate-cost equipment having the above outlined 
identifying and selecting capabilities has made it possible to express the 
essential contents of scientific and technical papers in the form of encoded 
abstracts, whose syntax is a simplified standardized system for recording 
relationships and whose component terms are encoded to express generic 
aspects of meaning 14 . Such a system can be applied to a broad field such 
as metallurgy. The more general application of this type of methodology 
opens up the possibility of establishing a network of agencies that would 
analyze, abstract and encode information in different fields on the basis 
of methodology based on common logical principles. Such a network— 
resembling somewhat the present world-wide telephone network—would 
greatly facilitate the interdisciplinary exchange of information and it 
would thus open up new prospectives for efficient use of scientific and 
technical information 16 . 

Principles of Code Development 

As already noted, the operations performed by, or with the aid of, 
various mechanical and electrical devices can be formulated by abstract 
expressions that are mathematical in nature. These operations, when so 
formulated, are, of themselves, as devoid of meaning as any other system 
of symbols or signs. They achieve meaning as a consequence of deliberate 
assignment of significance. This step is, rather obviously, of key impor- 

14 See footnote 7, page 399. 

14 For a more detailed discussion, see: J. W. Perry, Allen Kent and M. M. Berry, 
“Machine Literature Searching,” Chapter 8, New York, Iuterscience Publishers, 
Inc., 1956; also, J. W. Perry and Allen Kent “Documentation and Information 
Retrieval,” Chapter 3, New York, Interscience Publishers, Inc., 1957. 



SUBJECT MATTER ANALYSIS AND CODING 


407 


tance in determining the degree of benefit in applying mechanical and 
electrical devices to the searching, selection and correlation of recorded 
information. 

Appropriate assignment of meaning to certain forms of recording, e.g., 
the punches in cards, or the magnetic spots on computer tapes, does not, 
of course, accomplish the analysis of the subject contents of documents. 
Such analysis is equally necessary as a prerequisite to successful application 
of mechanical or electrical devices. Consistency in such analysis is, rather 
obviously, not assured by the fact that the selecting operations performed 
with the aid of various devices may be precisely formulated in terms of 
the Boolean notation. 

Appropriate assignment of meaning to certain forms of recording, i.e., 
the development of a specific code, must take into account, on the one 
hand, the nature of the selecting operations that can be performed by 
various types of devices or equipment and, on the other hand, the scope 
and the nature of the subject matter to be analyzed and encoded. 

Let us next direct attention to certain important consequences of the 
nature of the selecting operations that can be performed with various 
devices. 

Many different coding methods and different systems for applying 
various mechanical and electrical devices, as described in this book and 
elsewhere, make extensive use of selection based on the logical product. 
There is a fundamental reason for this which may be demonstrated as 
follows. 

Let us assume, for simplicity, that the subject contents of a file of 
documents can be analyzed in terms of 100 characteristics of which, on 
an average, ten will apply with equal probability to any one individual 
document. Then, a search for all documents pertaining to one characteristic, 
e.g., recorded as A, will result in one-tenth of the documents being selected. 
If, now, a search is directed to all documents pertaining to both of two 
subjects, e.g., to the logical product of the recordings A B, then one- 
hundredth of the documents will be selected. For a three-term logical 
product, e.g., A B C, one-thousandth, i.e., 0.1 per cent of the file will be 
selected. If we remember that we have assumed that the analysis of the 
subject contents of documents results in each code characteristic, recorded 
as A, B, C, D, etc., corresponding precisely to what the searcher would 
like to find under that heading, we may summarize this ideally favorable 
situation as shown in Table 18-1. Evidently searches defined in terms of 
logical products can permit us to achieve elimination of rapidly increasing 
proportions of the original file as the number of terms in the logical product 
is increased. In achieving this purpose with maximum efficiency, it is 
important that the terms in the logical product shall partake of the nature 
of independent variables. When this condition is not fulfilled, as in an 



408 


PUNCHED CARDS 


Table 18-1. Selectivity Increase with Incbeasb in Terms in Looical Product 
(Assuming Perfect Correspondence between Analysis and Search Requirement) 


Number of 



Per cent of Perti¬ 

Per cent of 

Terms in 



nent Information 

Information 

Logical 

Per cent of Total 

Per cent of Pertinent 

Lost by 

Selected that is 

Product 

File Selected 

Information Selected 

Non-selection 

not Pertinent 

1 

10 

100 

0 

0 

2 

1 

100 

0 

0 

3 

0.1 

100 

0 

0 

4 

0.001 

100 

0 

0 

n 

100(0.1)» 

100 

0 

0 


extreme case when all items coded for A are also coded for B, it is clear 
that a search directed to A-B will be no more selective than a search 
directed to either A or B above. 

The preceding paragraph presents one reason why considerable careful 
thought must be devoted to selecting those characteristics that are to 
serve as the terms in logical products for searching and selecting purposes. 
In the discussion of logical products that follows immediately and pertains 
to Tables 18-1 through 18-6, it will be assumed that the characteristics are 
not interdependent. In other words, the fact that some one characteristic 
pertains to the subject contents of a given document is without influence 
on the probability that any other of the characteristics in our hypothetical 
code pertain to the subject contents of the same document. 

In selecting characteristics to construct a code, another feature of 
logical products must also be taken into account. This relates to the 
possibility that a search directed to one or more characteristics does not 
result in selection of precisely those documents that are of pertinent interest 
to a given information requirement. There are three important possibilities: 

(1) Such a search does not result in selection of all pertinent information, 
but no non-pertinent information is selected. 

(2) Such a search selects all pertinent information but it also selects 
additional information not of pertinent interest. 

(3) Such a search does not result in selection of all pertinent information 
but it also results in selection of additional information not of pertinent 
interest. 

In the first case, by way of example, let us return to our previous ex¬ 
ample and assume that for any one characteristic, such as those recorded 
by A, B, C, etc., 9 per cent of the documents in the file is selected rather 
than 10 per cent as before. In this case, we will assume that the 9 per cent 
selected is all pertinent and that the remaining 1 per cent corresponds to 
pertinent information not selected but overlooked and consequently 



SUBJECT MATTER ANALYSIS AND CODING 


409 


Table 18-2a. Increase in Loss of Information with Increase 
in Terms in the Logical Product 


(Assuming each characteristic retrieves 90% of the information it should) 


Number of Terms 
in Logical Product 

Per cent of Total 
File Selected 

Per cent of Pertinent 
Information Selected 

Per cent of Pertinent 
Information Lott by 
Non-selection 

Per cent of 
Information 
Selected that 
is not Pertinent 

i 

9 

90 

10 

0 

2 

0.81 

81 

19 

0 

3 

0.0729 

72.9 

27.1 

0 

4 

0.00656 

65.6 

34.4 

0 

n 

100(0.09)" 

-(SC 

/ 0.09 V 
100 - 100 J 

0 


Table 18-2b. Increase in Loss of Information with Increase 
in Terms in the Logical Product 


(Assuming each characteristic retrieves 99% of the information it should) 


Number of Terms 
in Logical Product 

Per cent of Total 
File Selected 

Per cent of Pertinent 
Information Selected 

Per cent of Pertinent 
Information Lost by 
Non-selection 

Per cent of 
Information 
Selected that 
is not Pertinent 

1 

9.9 

99 

1 

0 

2 


98 

2 

0 

3 


97 

3 

0 

4 


96 

4 

0 

n 

100(0.99)" 


/0.099V 

.0° - 100 (—) 

0 


effectively lost. Under these conditions, the results of searches involving 
logical products with increasing numbers of terms may be summarized as 
shown in Table 18-2a. If, for each characteristic in the logical product, 
9.9 per cent of the file were selected (instead of 9 per cent) and the infor¬ 
mation lost by non-selection for any one characteristic were 0.1 per cent 
(instead of 1 per cent), the results would be more favorable as shown in 
Table 18-2b. 

As mentioned above, it is also possible for a search formulated by a 
logical product to select all pertinent information and, in addition, informa¬ 
tion that is not of pertinent interest. Let us assume that each term in the 
logical product selects 11 per cent of the file, with 10 per cent of pertinent 




410 


PUNCHED CARDS 


Table 18-3a. Increase in Extraneous Information with Increase in Terms 

in the Logical Product 


(Assuming each characteristic retrieves 11% of the 
file—10% pertinent, 1% extraneous) 


Number of 
Terms 
in Logical 
Product 

Per cent of Total 
File Selected 

Per cent of 
Pertinent 
Information 
Selected 

Per cent of 
Pertinent 
Information 
Lost by 
Non-selection 

Per cent of Information Selected 
that is not Pertinent 

i 

11 

100 

0 

9.1 

2 

1.21 

100 

0 

17.4 

3 

0.133 

100 

0 

24.9 

4 

0.0146 

100 

0 

31.7 

n 

100(0.11)" 

100 

0 

no.u,. - ( 0 .D .-1 

L (0.1D" J 


Table 18-3b. Increase in Extraneous Information with Increase 
in Terms in Logical Product 


(Assuming each characteristic retrieves 10.1% of the 
file—10% pertinent, 0.1% extraneous) 


Number of 
Terms in 
Logical 
Product 

Per cent of Total 
File Selected 

Per cent of 
Pertinent 
Information 
Selected 

Per cent of 
Pertinent 
Information 
Lost by 
Non-selection 

Per cent of Information Selected 
that is not Pertinent 

1 

10.1 

100 

0 

0.99 

2 

1.02 

100 

0 

1.96 

3 

0.103 

100 

0 

3.00 

4 

0.0104 

100 

0 

3.86 

n 

100(0.101)" 

100 

0 

T(0101)" - (0.1)"“1 

L (0101)" J 


interest and 1 per cent extraneous and non-pertinent. In this case, results 
as shown in Table 18-3a would be obtained. If, for each characteristic in 
the logical product, 10.1 per cent of the file were selected (with 10 per cent 
pertinent information and 0.1 per cent not pertinent), then the results, as 
shown in Table 18-3b would be more favorable. 

It is, of course, quite possible that a selecting operation directed to a 
given characteristic may result in selection of extraneous information 
while, simultaneously, all pertinent information is not retrieved. For 



SUBJECT MATTER ANALYSIS AND CODING 


411 


Table 18-4a. Increase both in Lost Information and in Extraneous 
Information with Increase in Terms in Logical Product 


Number of 
Terms in 
Logical 
Product' 

Per cent of 
Total File 
Selected 

Per cent of 
Pertinent 
Information 
Selected 

Per cent of Pertinent 
Information Lost 
by Non-selection 

Per cent of Information 
Selected that is not Pertinent 

1 

10 

90 

10 

10 

2 

i 

81 

19 

19 

3 

0.1 

72.9 

27.1 

27.1 

4 

0.01 

65.6 

34.4 

34.4 

II 

100(0.1)“ 

"O' 

/0.09\“ 

.0° 

* r<°*>- - ®»-i 

L (o.i)» J 


Table 18-4b. Increase both in Lost Information and in Extraneous 
Information with Increase in Terms in Logical Product 


Number of 
Terms in 
Logical 
Product 

Per cent of 
Total File 
Selected 

Per cent of 
Pertinent 
Information 
Selected 

Per cent of Pertinent 
Information Lost 
by Non-selection 

Per cent of Information 
Selected that is not Pertinent 

i 

10 

99 

1 

i 

2 

1 

98 

2 

2 

3 

0.1 

97 

3 

3 

4 

0.01 

96 

4 

4 

n 

100(0.1)" 

100(0.099)“ 

/0.099\" 

100 - 100 ( ) 

V 0.1 ) 

T (0- 1 )" - (0.099)"”j 

L (0-D* J 


example, let us assume that a search of any one characteristic results in 
selecting 10 per cent of the file of which 10 per cent of the pertinent in¬ 
formation constitutes 9/10 and the non-pertinent information 1/10. In this 
case, increasing the number of terms in the logical product leads to the 
results shown in Table 18-4a. If, in the case under consideration, the pro¬ 
portion of pertinent information selected by any one characteristic in¬ 
creases from 9/10 to 99/100 and, correspondingly, the proportion of 
extraneous information decreases from 1/10 to 1/100, then much more 
favorable results are obtained as shown in Table 18-4b. 

In working out numerical examples to illustrate the various effects of 
increase in number of terms in a logical product on search results, the 
calculations have been kept simple by assuming that for each of the special 
cases investigated, all the characteristics used to analyze the subject 




412 


PUNCHED CARDS 


n = totality of documents em¬ 
braced by the system 
m = documents to which sys¬ 
tem directed attention 
w (cross hatched) = documents 
found on inspection of "m" 
to be of pertinent interest 
x ~ documents of actual per¬ 
tinent interest 


Figure 18-1. Diagram exemplifying general case of relationship between n, m, x 
and w. 

contents of documents are functionally effective in the same way and to 
the same degree. Thus in the calculations for Table l8-4a, it was assumed 
that each characteristic resulted in selection of 10 per cent of the file and 
that 9/10 of the selected items would be of pertinent interest. In actual 
situations, such uniformity of functional effectiveness of all characteristics 
can scarcely be anticipated. If we assume differences, however, the character 
of the results remains unchanged. This is true, in particular, of the trends 
observed as the number of terms in the logical product are increased. To 
illustrate this point, the diagram shown in Figure 18-1 is helpful. Here the 
area within the circle n indicates the totality of documents in a given 
file while the area of the circle m indicates those documents which are 
selected by a search directed to a given characteristic, as recorded by 
code A. The area of circle x indicates those documents that a person 
making use of the system would like to have brought to his attention by 
a search directed to the encoded characteristic A. The shaded area, W, 
indicates those documents of pertinent interest that are included among 
the documents selected as encoded for a given characteristic. The circles 
m and x will be congruent, if, and only if, there is complete agreement 
between the person who does the encoding and the person who needs 
information as to which documents should be characterized by the encoded 
characteristic A. In general, the degree of agreement will deviate, more or 
less, from 100 per cent and various cases of deviations from complete 
agreement have been discussed elsewhere 16 . 

In previous discussion, it was emphasized that the areas within the 
various circles in diagrams such as Figure 18-1 and also the degree of 
overlap of circles m and x are independent variables, which are of decisive 
importance in determining the operational effectiveness of an information 
system when it is used to select documents from a file or library. To il¬ 
lustrate this important point further, let us assume that we have an 




SUBJECT MATTER ANALYSIS AND CODING 


413 


Table 18-5a. Data to Illustrate Search Results 
with Different Characteristics 


CharatUrislic 

• 

m 

X 

V 

A 

15000 

300 

250 

240 

B 

15000 

1500 

1515 

1490 

C 

15000 

750 

1000 

700 

D 

15000 

160 

156 

140 

Table 18-5b. 

Operational Search 

Factors with 

Different Characteristics 

Characteristic 

m/n 

x/n 

w/x 

w/m 

A 

2% 

1.667% 

96.0% 

80% 

B 

10% 

10.1% 

98.35% 

99.4% 

C 

5% 

6.67% 

70% 

93.4% 

D 

1.07% 

1.04% 

89.75% 

87.5% 


information system in which four characteristics perform, in a qualitative 
fashion for various interrelated searches, as shown in Figure 18-1 and that 
the quantitative performance provides data as given in Table 18-5a. The 
corresponding values for m/n, x/n, w/m and w/x are given in Table 18-5b 
for the four subject matter characteristics A, B, C, D. When logical products 
are set up among these four characteristics to define the scope of searches, 
changes must be anticipated in the per cent of pertinent information 
retrieved (and also the per cent of pertinent information lost by non¬ 
selection) as well as in the per cent of selected information that is pertinent 
(and also the per cent of selected information that is not pertinent). These 
changes, as shown in Table 18-6 were calculated in the same way as the 
data presented